A Closer Look at Deep Learning On Tabular Data
A Closer Look at Deep Learning On Tabular Data
Han-Jia Ye1 Si-Yang Liu1∗ Hao-Run Cai1∗ Qi-Le Zhou1 De-Chuan Zhan1
1
School of Artificial Intelligence, Nanjing University, China
National Key Laboratory for Novel Software Technology, Nanjing University
{yehj,zhouql,zhandc}@lamda.nju.edu.cn,{liusiyang,caihr}@smail.nju.edu.cn
arXiv:2407.00956v1 [cs.LG] 1 Jul 2024
Abstract
1 Introduction
Machine learning has achieved success in a wide range of domains. Tabular data, where datasets
are organized in table format, are among the most commonly used data types in machine learning,
e.g., CTR prediction [30, 55], healthcare [24], medical analysis [44, 48], and e-commerce [39]. In
a tabular dataset, each row represents an instance, and each column corresponds to an attribute
(aka. a feature). In supervised learning, a training instance is associated with a label (discrete for
classification tasks and continuous for regression tasks). A machine learning model learns a mapping
from an instance to its label on the training set and generalizes this ability to unseen test instances
sampled from the same distribution.
Machine learning methods for tabular datasets are constantly evolving. Classical methods like Logistic
Regression (LR), Support Vector Machine (SVM), Multi-Layer Perceptron (MLP), and decision tree
have been proposed long ago and serve as the basis for a wide range of algorithms [7, 37]. For practical
usage, tree-based ensemble methods have demonstrated their advantages in many applications [13,
∗
These authors contribute equally to this work.
Time
Time
SVM XGBoost SVM
100 XGBoost 100 KNN
Random Forest KNN Random Forest
10−1 LR 10−1 LR
Optimal Optimal
TabPFN Dummy TabPFN Dummy
Benchmark Regression
1 2 3 4 5 6 7 8 9 10 11 Full
12 Benchmark
13 1 2 3 4 5 6 7 8 9 10 11 12 13
Average Rank Average Rank
Time
Time
SVM
100 XGBoost 100
XGBoost KNN
KNN
10−1 Random Forest 10−1 Random Forest
LR
Optimal LR Optimal
Dummy Dummy
1 2 3 4 5 6 7 8 9 10 11 12 13 1 2 3 4 5 6 7 8 9 10 11 12 13
Average Rank Average Rank
33, 41]. In recent years, Deep Neural Networks (DNN) have been successfully applied in vision
and language domains [46, 50, 17]. Inspired by this success, attribute embeddings [47, 19], data
augmentation [56], and deep models for tabular data have been designed [8, 57, 58, 60, 61]. Early
deep tabular methods struggled to achieve good performance due to their high complexity, leading to
efforts to capture complicated feature correlation and mimic the workflow of tree-based models [14,
22, 40, 10]. However, further studies have shown that modern deep learning techniques [4, 18] enable
classical methods such as MLPs to excel [31].
Considering several advantages of deep tabular models beyond performance, such as their ability to
capture complex feature interactions via nonlinear mappings and their flexibility to adapt to special
requirements, evaluating these models is essential for their development. Unlike the visual and
textual domains, which have widely accepted benchmarks, datasets for evaluating tabular methods
are collected from various sources such as UCI [23], OpenML [49], and Kaggle. Although there
have constructed some tabular benchmarks [8, 21, 36], they are limited in terms of the coverage
of evaluated methods, dataset sizes, and types of tabular tasks. Furthermore, the dilemma between
the need for many tabular datasets to thoroughly reveal the ability of a given method and the high
computational burden makes a large benchmark challenging for practical usage.
In this paper, we first construct a comprehensive benchmark comprising 300 tabular datasets, covering
binary classification, multi-class classification, and regression tasks from diverse domains. Datasets
in our benchmark are distributed from small to large sizes in a more uniform manner. We perform an
extensive comparison between deep tabular methods and tree-based methods, revealing the average
rank of all methods fairly. Our results provide insights into applying tabular methods based on the
intrinsic properties of both methods and datasets.
We also record the training dynamics, including changes in losses and other statistics for the training,
validation, and test stages, along with the training progress for each dataset-method pair. Instead of
merely analyzing the final tabular results, we propose a new tabular prediction task: predicting the
evolution of a performance curve based on its initial points. Specifically, the model learns a mapping
from the meta-features of datasets and the initial part of the dynamics, to the validation set statistics
over epochs. This model helps extract essential meta-features that influence prediction accuracy and
emphasizes the key factors influencing task preference for different types of methods.
2
Based on the full evaluations of our large benchmark, we identify two subsets of 45 datasets. The
first subset contains datasets that favor either tree-based methods or DNN-based methods, serving as
an effective analysis tool to check whether a strategy benefits deep tabular models. We validate this
subset by investigating when an attribute encoding strategy is beneficial. The second subset contains
datasets where the method ranks are consistent with the overall benchmark, acting as a probe for
tabular analysis. These “tiny tabular benchmarks” will facilitate further studies on tabular data. The
contributions of our paper are:
• We propose a large tabular classification and regression benchmark and evaluate up-to-date tabular
methods comprehensively.
• We propose a novel task predicting the training dynamics (the validation performance curves) from
the initial points and the meta-features of a dataset. The learned model helps tabular data analysis.
• Two tiny benchmarks with 15% of the whole size are also extracted to facilitate further tabular
research in a lightweight manner.
2 Preliminary
2.1 Learning with Tabular Data
ℓ(·, ·) measures the discrepancy between the predicted label ŷi and the true label yi , e.g., cross-entropy
in classification. Ω(·) is the regularization on the model. We expect the learned f is able to extend its
ability to unseen instances sampled from the same distribution as D.
We consider several types of tabular models to implement f , the classical methods, the tree-based
methods, and deep tabular methods. Classical methods contain LR, KNN, and SVM. We also
consider the dummy baseline, which outputs the label of the major class and the average labels for
classification and regression tasks, respectively. For tree-based methods, we consider Random Forest,
XGBoost[13], CatBoost [41], and LightGBM [33]. Based on the workflow of deep tabular methods,
we categorize them into three main groups: those that focus on handling input features, those that
design the learning objectives, and those that apply specific learning algorithms.
Raw-feature-based Methods. Similar to the vanilla MLP working on raw features, many deep
tabular methods adopt the same strategy, but with different model architectures for the mapping f .
In addition to the newly designed MLP [18], architectures such as Residual Network (ResNet)[18],
Self-Normalizing Neural Networks (SNN)[34], DANets [11], TabCaps [12], and DCNv2 [53] are
also applied. Tree-mimic Methods design their architectures to mimic the decision strategy of
trees, including NODE [40], GrowNet [5], and TabNet [4]. Unlike parametric methods, some
Neighborhood-based Methods use non-parametric strategies to predict the label of a target instance
based on the entire training set. Examples include DNNR [38] and TabR [20].
Instead of applying the model directly to raw features, Token-based Methods transform feature
vectors into a set of tokens. Each feature value is mapped to a high-dimensional vector using a
lookup table or linear mapping. These learnable tokens facilitate the final predictions. Representative
methods include AutoInt [47], TabTransformer [28], and FT-Transformer (FT-T) [18].
Regularization-based Methods regularize the predictions of deep tabular models, e.g., TAN-
GOS [29] and SwitchTab [54]. PTaRL is a prototype-based method based on space calibration [59].
3
140 Ours 60 Ours 60 Ours
120 TabZilla 50 TabZilla TabZilla
100 40
Count
40
80
Count
Count
30 20
60
40 20
0
20 10
Ed mp on
Biohysical
He log al
h l
Ind atur e
u al
cie l
Co Visi al
ati r
Ha Le onal
cra re
Ot fted
rs
alt ica
So stria
uc ute
N car
P nci
he
nd isu
a
0 0 0.0
Fin
Binary Multi-Class Regression 0.2 0.4 0.6 0.8 1.0
Task Type Size 1e6 Domain
(a) Statistics over task types (b) Statistics over data sizes (c) Statistics over domains
Figure 2: Advantages of the proposed benchmark. (a) shows the number of datasets for three tabular
prediction tasks. (b) shows the number of datasets along the change of their sizes (N × d). (c) shows
the histogram of datasets across various domains.
General Deep Methods pre-train a deep neural network can be applied to any downstream tabular
task, typically without requiring additional hyper-parameter tuning [27, 45, 62]. Specifically, we
use TabPFN [27], where pre-trained transformers handle small-size tabular classification tasks.
Following [36], we randomly sample 3000 examples when the training data size is large.
We propose a large tabular benchmark for comprehensive evaluation. First, we describe the steps of
constructing this benchmark as well as the advantages, followed by the comparison results.
Datasets. The datasets in our benchmark are collected from UCI [23], OpenML [49], and Kaggle.
We mainly filter and pre-process datasets as follows:
• Size selection. A dataset is considered if N > 1000 and d > 5, since too small datasets are difficult
to evaluate due to their limited test sets.
• Missing value ratio. We exclude datasets with more than 20% missing values.
• Outlier dataset. Classification datasets are excluded if they are too imbalanced (where an MLP
predicts all instances as the major class) or too easy (where an MLP achieves over 99% accuracy).
• Attribute pre-processing. Given a dataset, we delete “id/index/timestamp” attributes. We use
ordinal encoding for categorical features [8, 36]. Then we follow [18] to pre-process the dataset.
Since there exist duplicate datasets in UCI and OpenML, we neglect copies of a dataset as well
as those smaller subsets sampled from a larger one. Ultimately, we have 101 binary classification
datasets, 80 multi-class classification datasets, and 119 regression datasets. Details are in Appendix A.
Implementation Details. Given that the performance of tabular methods depends on tuned hyper-
parameters, we follow setups in [18, 20] to evaluate all methods. Each dataset is randomly split into
training/validation/test partitions with proportions of 64%/16%/20%, respectively. Hyper-parameters
are tuned and early stopping is performed on the validation set. All hyper-parameters are searched
using Optuna [2] over 100 trials. The best-performing hyper-parameters are then used to train the
model with 15 random seeds, and the average performance is reported. For all deep methods, we set
the batch size to 1024 and use AdamW [35] as the optimizer.
Evaluation. We use accuracy for classification (higher is better) and Root Mean Square Error
(RMSE) for regression (lower is better) to select the best-performing model during training on the
validation set. Additionally, we consider F1/AUC for classification and MAE/R2 for regression to
evaluate test set performance. Following [16, 36], we report the average performance rank among all
methods and datasets (lower is better).
Comparison Methods. We compare all methods mentioned in subsection 2.2, with full results
included in Appendix D. Representative methods selected for analysis in the main paper, i.e., LR,
KNN, SVM, XGBoost, CatBoost, RandomForest, MLP, ResNet, FT-T, DCNv2, TabR, and TabPFN.
4
Table 1: The comparison results of all methods on our benchmark. We list the detailed results on
binary classification, multi-class classification, and regression tasks with the default hyper-parameters
and the tuned ones. The “min”, “max”, and “avg.” denote the minimum, maximum, and average
rank among all methods over datasets belonging to a certain tabular task. “Dummy” means the naive
baseline. “LR”, “XGB”, and “CatB” denote “Logistic regression/Linear Regression”, “XGBoost”,
and “CatBoost”, respectively. TabPFN cannot be applied to regression tasks.
Binary Classification Multi-Class Classification Regression
Default Tuned Default Tuned Default Tuned
min max avg. min max avg. min max avg. min max avg. min max avg. min max avg.
Dummy 1 13 11.97 1 13 11.93 1 13 12.31 2 13 12.30 1 12 11.24 1 12 11.26
LR 1 13 8.90 1 13 8.03 1 12 9.00 1 12 8.54 2 12 9.18 1 12 8.59
KNN 2 13 10.32 1 13 8.91 1 13 9.36 1 13 7.83 1 12 8.73 1 11 7.13
SVM 1 13 7.94 1 13 8.54 1 12 8.06 1 13 9.78 1 12 8.07 1 12 9.63
XGB 1 13 5.65 1 13 4.77 1 12 5.45 1 12 5.26 1 11 5.44 1 12 4.13
CatB 1 11 3.67 1 12 4.24 1 11 3.91 1 12 4.44 1 8 2.41 1 11 3.15
RForest 1 13 5.59 1 13 6.10 1 12 5.74 1 12 6.42 1 11 5.08 1 12 5.24
MLP 1 11 6.33 1 13 6.61 1 12 6.88 1 12 5.90 1 11 5.86 1 12 6.25
ResNet 1 13 6.42 1 13 6.85 1 12 5.20 1 12 5.37 1 10 5.50 1 12 6.23
FT-T 1 12 5.62 1 13 5.58 1 12 5.28 1 12 5.50 1 12 4.89 1 12 5.02
DCNv2 1 13 6.09 1 13 6.98 1 13 6.44 1 13 6.15 1 12 6.52 1 12 6.28
TabR 1 13 5.03 1 13 4.70 1 11 3.99 1 13 3.79 1 12 4.60 1 12 4.97
TabPFN 1 12 6.26 1 12 6.36 1 11 6.73 1 12 6.92 - - - - - -
3
DNN-Friendly 12 LR MLP TabR 12.5 LR MLP TabR
2 Tree-Friendly CatB FT-T CatB FT-T
10 10.0
1
Rank
8 7.5
Score
Rank
0 5.0
6
1 2.5
4
2
Ed ompsion
B hys ial
Heiological
h l
Ind atu re
us ral
cie l
C Vi tal
at er
nd isu l
cra re
Ot fted
rs
alt ica
So tria
Ha Leiona
he
N ca
uc ut
2 P nc
a
Fin
3 0 50 100 150 200 250 300 ~1e4 ~1e5 ~1e6 ~1e7 ~1e8
Datasets Size Domain
(a) Tree vs. DNN methods (b) Comparison across data sizes (c) Comparison across domains
Figure 3: Comparison results on our benchmark. The score in (a) is calculated as the performance
difference between the best tree-based method and the best DNN-based method. Average performance
rank is used as a criterion in (b) and (c).
Our proposed benchmark has several advantages on evaluating and analyzing tabular methods
compared with the previous one, TabZilla [36].
Coverage of Tasks. Unlike [36], which only considers classification tasks, our benchmark also
includes 119 regression tasks. Evaluations on regression datasets are crucial since most deep tabular
methods can handle both classification and regression simultaneously. The numbers of datasets for
binary classification, multi-class classification, and regression tasks are illustrated in Figure 2 (a).
Coverage of Data Sizes. Our benchmark covers a wide range of datasets with diverse sizes,
particularly in terms of N and d. Although TabZilla includes datasets with a wide range of N , most
datasets are either too small or extremely large. As shown in Figure 2 (b), our benchmark offers a
more uniform distribution of dataset sizes w.r.t. N × d.
Coverage of Domains. We collect datasets from 13 domains, including healthcare, biology, finance,
education, and physics. This diverse collection allows us to better assess whether a tabular method
can generalize across different applications. Details are shown in Figure 2 (c).
The comparison results of 13 representative methods, including those tuned models and models with
default hyper-parameters, on the proposed benchmark are listed in Table 1. In addition, Figure 1
demonstrates the performance-efficiency-size comparison of these tabular methods on our benchmark,
5
with classification tasks shown in (a) and (b), regression tasks in (c), and all task types in (d). Based
on the results, we have several observations:
• CatBoost achieves the best average rank in most classification and regression tasks, which is
consistent with [36]. Among all deep tabular methods, TabR works the best in most cases. However,
it has a high training cost, as demonstrated in Figure 1.
• Most methods benefit significantly from hyper-parameter tuning. Since the average rank reveals
the relative comparison between methods, a lower rank after hyper-parameter tuning indicates
substantial improvements for certain methods (e.g., classical methods like LR and deep methods
like TabR). Although the average ranks of some tuned models, such as CatBoost, may increase, this
suggests that while their performance improves, they are still surpassed by other methods on certain
datasets. In other words, their relative improvements are smaller compared to other methods.
• For deep tabular methods, the variance in their ranks between classification and regression tasks
is usually stable. Moreover, the ranks of deep tabular methods improve (i.e., become lower)
when extending the same method from regression to binary classification and then to multi-class
classification tasks. Since some classical linear methods perform well on binary classification tasks,
the ranks of deep methods may be slightly higher in this case.
• We calculate the difference in performance between the best tree-based methods and the best
DNN-based methods. We sort the differences and present the results in Figure 3 (a), where higher
scores indicate that the dataset is more tree-friendly. The varied results between tree-based and
DNN-based methods suggest that datasets have a preference for specific types of methods.
• Nonlinear models archive better results in most cases than linear ones. However, there indeed exist
several datasets where the naive baseline and linear model (such as LR) achieves the best results.
• Figure 3 (b) shows the results on different sizes (N × d) ranging from 1e4 to 1e8. We find that
TabR works better than other methods on smaller-sized datasets, while CatBoost maintains its
superiority on larger datasets.
• We also demonstrate the results of different methods across diverse domains in Figure 3 (c). For
example, CatBoost outperforms the second-best method by a large margin on industrial, societal,
and educational datasets.
In addition to performance measures of different methods, our benchmark also records detailed
statistics during the training process. Specifically, given a training set D of a dataset, we optimize a
deep tabular model f by stochastic gradient descent over Equation 1. In each epoch, we randomly
permute the order of all examples in D, and sample mini-batches of examples sequentially. An epoch
is defined as one complete pass through all examples in D. We can record the change in the validation
set statistics (e.g., accuracy for a classification task) in the form a = [a1 , a2 , . . . , aT ] ∈ RT+ over T
epochs before early stopping.
Training deep tabular models usually incurs high time and computational costs. Therefore, an efficient
way to estimate the model’s ability is to predict subsequent values in a given the initial values. In
other words, we aim to predict the dynamics of validation set statistics curves based on their initial
parts. If the accuracy curve for a classification task does not increase rapidly but oscillates, it is
reasonable to stop the training and try another configuration of the model [15]. To this end, we
propose a new tabular prediction task: predicting the dynamics of a model during its training progress
based on both the properties of D as well as the initial values of the curve a.
In particular, we collect meta-features mD for D, which include basic statistics of a dataset (N , d,
and C), and other criteria such as the joint entropy between each attribute and class [36]. Additionally,
T −K
we define a support set S ∈ RK + containing the first K values in a, and a query set Q ∈ R+
consisting of the remaining values. We aim to learn a mapping g from {mD , S} to Q, based on
the statistics of all datasets and methods in our benchmark. We set a as the list of accuracies for
classification and as the list of normalized RMSE for regression tasks. The support set S depicts the
6
1.0 1.0 1.0
Train Data Train Data
0.70 Test Data Test Data
0.9 0.9 0.9
Fit Fit
0.8 0.8 Predict Predict
0.8
Valid Accuracy
Valid Accuracy
0.65 Scaling law Scaling law
Valid RMSE
Valid RMSE
0.7 0.7
0.7
0.60 0.6 0.6
0.6
Train Data Train Data 0.5
0.5
Test Data Test Data 0.4 0.5
0.55
Fit 0.4 Fit
Predict Predict 0.3 0.4
Scaling law 0.3 Scaling law
0.50 0.2
0 10 20 30 40 50 60 0 20 40 60 80 100 120 140 0 25 50 75 100 125 150 175 200 0 25 50 75 100 125 150 175 200
Epoch Epoch Epoch Epoch
Valid Accuracy
Valid Accuracy
Valid RMSE
Valid RMSE
0.80 0.93 0.99 0.6
0.78 0.92 0.98
0.97 0.4
Train Data 0.91 Train Data
0.76 Test Data Test Data 0.96
Fit 0.90 Fit 0.2
0.74 Predict Predict 0.95
Scaling law 0.89 Scaling law
0.94
0 20 40 60 80 100 120 0 10 20 30 40 50 0 10 20 30 40 50 60 70 0 25 50 75 100 125 150 175 200
Epoch Epoch Epoch Epoch
Figure 4: The visualization results of fitting the training dynamics (the validation curves of an MLP
trained with default hyper-parameters) on eight datasets are shown. The datasets in the first and
second rows are DNN-friendly and Tree-friendly, respectively. The first two columns represent
classification tasks, while the second two columns represent regression tasks.
initial changes in the validation curve, while the meta-features mD encode the intrinsic properties
of a dataset. By learning g, we can analyze which meta-features influence changes in validation set
accuracy, thereby identifying the key factors essential for the performance of deep tabular methods.
Given the support set S, one direct strategy to fit and extrapolate the curve is to apply the neural
scaling law [25, 42, 6, 26, 3, 1, 9]. The scaling law defines a curve family with several parameters,
which can be estimated based on S. Due to the limited values in S, we propose a baseline for this
task via meta-learning a mapping from S to the parameters across various datasets.
We use the following function family to depict the curve:
√
aθ (t) = A log t + B t + C + D/t . (2)
Here t is an integer representing the epoch number, and a(t) outputs the validation set statistics for a
given epoch index. θ = {A, B, C, D} are the parameters of the curve. The variability of the curve is
determined by θ, so we aim to learn a meta-mapping h from {mD , S} to θ [51].
Although the curve may be predicted based only on the support set S, we include the meta-features
mD as an additional input source. mD not only provides auxiliary information about the dataset but
also makes the predictions depend on the properties of the dataset. The learning objective of h is
X X
min ℓ(aθ=h(mD ,S) (t), at ) . (3)
h
{mD ,S} at ∈{Q}
We randomly sample a (batch of) dataset from the benchmark, and for each dataset, we collect mD ,
S, and Q. The model predicts the parameter θ of the curve family, and the quality of the prediction is
based on the discrepancy (measured by ℓ) between the predicted values via θ and the true values in Q.
We evaluate the ability of h given an unseen dataset.
We analyze tabular methods by fitting the mapping h in four scenarios: classification vs. regression
and tree-friendly vs. DNN-friendly datasets. Details on splitting datasets are in the next section.
The visualization results of fitting the training dynamics (the validation curves of an MLP trained with
default hyper-parameters) on eight datasets are shown in Figure 4. For more details, see Appendix B.
The datasets in the first and second rows are DNN-friendly and tree-friendly, respectively. The first
and second columns represent classification and regression tasks. Our proposed method fits the curve
more accurately than using the scaling law directly. Besides, the inclusion of meta-features improves
the fit in most cases. The results indicate that our method effectively estimates the training dynamics
7
NN Regression
10 1 NN Classification Greedy Random Clustering
Feature Importance
Tree Regression
Tree Classification Seen Unseen Seen Unseen Seen Unseen
Split 1 0.85 1.17 0.16 0.11 0.47 0.62
Split 2 1.07 1.92 0.16 0.58 0.49 0.66
10 2
Split 3 1.48 1.63 0.18 0.26 0.49 0.78
Split 4 0.90 1.85 0.19 0.40 0.47 0.70
ity ttr
co st
ge tio
ran attr
cla nr an
t_t an
im ean. vity
nt_ an
co t.sd
sp nr an
ran .sd
m sd
r_e d
n t.sd
me .sd
.sd
mu nce an
nf o
_ou n
rs
t_i rati
att ax.s
nr .mea
ss_ _in
tlie
Split 5 1.03 1.43 0.13 0.47 0.55 0.52
.
e
ins .me
joi c.me
bla me
ars _a
ran s_ra
iq_ nc
ge
ge
an
.m
v.m
m ra
o_
en
g
n
co
ss_
Meta Feature
Figure 5: The importance of meta-features Table 2: MAE of the ranks of methods on the
when we train a predictor for training dynam- proposed tiny benchmark and the ranks on the
ics over all the datasets. whole benchmark.
of a deep tabular model. Based on these estimates, we can intelligently choose the best configurations
of models by early stopping those unlikely to achieve good performance.
Combining the curve fitting results in Figure 4 with the feature importance results in Figure 5,
we find that meta-features with large weights are meaningful, revealing that the complexity of the
dataset matters in predicting the training dynamics. For classification tasks, the distance to the center
(gravity) is important for fitting the curve for both tree-based and DNN-based methods. For regression
tasks, the range of attributes (range.mean) and the mean of attributes (mean.mean) are essential. For
tree-based methods, the sparsity (sparsity.mean) of attributes is crucial, while DNN-based methods
pay more attention to the statistics of attributes (such as max.sd).
The different behaviors of tree-based tabular methods and DNN-based methods have been observed
and analyzed in previous studies [21, 36] and in section 3. In other words, tree-based and DNN-based
methods have different preferences for datasets. We construct a tiny benchmark where the difference
between the two types of methods is pronounced, which will be useful for tree-vs.-DNN analysis.
Selection Strategy. We select a subset of datasets where tree-based and DNN-based methods exhibit
diverse behaviors based on their performance ranks in the full benchmark. Specifically, for a given
dataset and the performance criteria of different methods, we define a dataset as “tree-friendly (TF)” if
tree-based approaches (i.e., RandomForest, XGBoost, CatBoost) achieve higher performance than the
best DNN-based methods (i.e., MLP, ResNet, FT-T, TabR). Conversely, a dataset is “DNN-friendly
(DF)” if DNN-based methods outperform tree-based ones. If the performance of the two types of
methods is similar (i.e., the difference between their results is smaller than a threshold), we classify the
dataset as a “tie”. We partition all datasets into several groups based on their sizes. Specifically, there
are 5, 4, and 6 groups for binary classification, multi-class classification, and regression, respectively.
For each subset, we choose one dataset from each of the three cases: “TF”, “DF”, and “Tie”.
Applications of the benchmark. The success of deep tabular models relies on some key modules,
but the proportion of their helpfulness also depends on the property of a dataset. Since the datasets in
this tiny benchmark have diverse preferences, we demonstrate the benchmark becomes a useful tool
for analyzing some special modules in deep tabular methods, e.g., attribute encoding strategies. In
addition to the vanilla strategy which keeps the original numerical value, several feature encoding
strategies have been proposed in previous studies for numerical feature xnum
ij , i.e., Piece-wise Linear
Encoding strategies (quantile version PLE-Q and tree-based version PLE-T) [19, 20]. We equip
representative tabular methods with these encoding strategies and report their results on this tiny
benchmark in Table 3. Although the PLE encoding generally benefits deep tabular methods, we
find that it is particularly helpful on tree-friendly datasets. However, on DNN-friendly datasets, it
sometimes leads to negative effects.
8
Table 3: Comparison results over the first proposed tiny benchmark. We use “TF”, “DF”, and “Tie” to
denote the subsets of datasets that are tree-friendly, DNN-friendly, and those without any preference,
respectively. Various encoding strategies are applied to the numerical features. In addition to the
vanilla one, we also investigate PLE-Q and PLE-T, denoted by “E-Q” and “E-T”, respectively.
MLP ResNet FT-T TabR
Vanilla w/ E-Q w/ E-T Vanilla w/ E-Q w/ E-T Vanilla w/ E-Q w/ E-T Vanilla w/ E-Q w /E-T
TF 2.25 1.92 1.83 1.92 2.08 2.00 2.33 1.75 1.83 2.08 1.83 2.08
Tie 1.83 2.00 2.17 1.50 2.33 2.17 1.67 2.08 2.17 1.75 2.33 1.92
DF 1.25 2.38 2.38 1.75 2.19 2.06 1.50 2.00 2.50 1.69 2.00 2.31
XGBoost RandomForest SVM Catboost
Vanilla w/ E-Q w/ E-T Vanilla w/ E-Q w/ E-T Vanilla w/ E-Q w/ E-T Vanilla w/ E-Q w/ E-T
TF 2.67 1.83 1.50 1.67 2.00 2.33 2.17 2.17 1.67 2.25 1.83 1.83
Tie 2.08 1.83 1.83 1.58 2.25 2.08 2.33 1.92 1.67 1.92 2.00 1.83
DF 2.18 1.94 1.88 2.19 1.88 1.94 2.25 1.94 1.75 2.25 1.81 1.88
5.2 Tiny Benchmark for Rank Consistent Evaluation
The large size of the full benchmark increases the burden of evaluating a new tabular method.
Therefore, we propose to extract a subset of the datasets and expect the average ranks of these tabular
methods on this tiny benchmark to be consistent with the ranks on the full benchmark.
Selection Strategy. Given a dataset D in the benchmark B, we can evaluate L methods A via
ranking(D, A) ∈ RL + . Based on the performance record over B, we formulate the selection process
as an optimization problem:
X X
min
′
ℓ( ranking(D, A), ranking(D′ , A)) s.t. |B ′ | ≤ ηD . (4)
B ⊆B
D∈B D ′ ∈B′
η is the ratio that controls the number of datasets selected from the benchmark, and ℓ measures the
difference between the average rank based on the selected benchmark B ′ and the whole benchmark B.
To solve the combinatorial optimization, we try three strategies: a greedy strategy, a random selection
over 10,000 trials, and a clustering method (KMeans) based on all rank results.
Applications of the benchmark. We validate whether this proposed tiny benchmark can produce
consistent ranks for a new method. Since the benchmark is selected based on 13 representative tabular
methods, we split these methods into five groups. We select subsets using four groups and test the
quality of the selection strategy on the remaining group. Specifically, we measure the MAE of a
given strategy between the ranks of the seen/unseen methods on the selected benchmark compared to
the ranks on the whole benchmark. We find that the random strategy obtains lower MAE on both
seen and unseen groups of methods in Table 2. Finally, we use the same strategy to select the tiny
benchmark based on the results of all 13 methods.
6 Conclusion
We provide a comprehensive evaluation of machine learning methods for tabular data. Our benchmark
includes a large number of datasets covering a wide range of domains, data sizes, and task types.
Beyond evaluating multiple performance criteria for different methods, we also record the training
dynamics, introducing a novel task: predicting the validation curve based on its initial part and the
dataset’s meta-features. Finally, we extract two subsets of tiny benchmarks. One subset reveals the
diverse properties of DNN and tree-based models, while the other maintains consistent ranks of
different methods with the full benchmark, facilitating further studies on tabular data.
References
[1] Samira Abnar, Mostafa Dehghani, Behnam Neyshabur, and Hanie Sedghi. Exploring the limits
of large scale pre-training. In ICLR, 2022.
[2] Takuya Akiba, Shotaro Sano, Toshihiko Yanase, Takeru Ohta, and Masanori Koyama. Optuna:
A next-generation hyperparameter optimization framework. In KDD, pages 2623–2631, 2019.
9
[3] Ibrahim M. Alabdulmohsin, Behnam Neyshabur, and Xiaohua Zhai. Revisiting neural scaling
laws in language and vision. In NeurIPS, pages 22300–22312, 2022.
[4] Sercan Ö. Arik and Tomas Pfister. Tabnet: Attentive interpretable tabular learning. In AAAI,
pages 6679–6687, 2021.
[5] Sarkhan Badirli, Xuanqing Liu, Zhengming Xing, Avradeep Bhowmik, and Sathiya S. Keerthi.
Gradient boosting neural networks: Grownet. CoRR, abs/2002.07971, 2020.
[6] Yasaman Bahri, Ethan Dyer, Jared Kaplan, Jaehoon Lee, and Utkarsh Sharma. Explaining
neural scaling laws. CoRR, abs/2102.06701, 2021.
[7] Christopher Bishop. Pattern recognition and machine learning. Springer, 2006.
[8] Vadim Borisov, Tobias Leemann, Kathrin Seßler, Johannes Haug, Martin Pawelczyk, and
Gjergji Kasneci. Deep neural networks and tabular data: A survey. IEEE Transactions on
Neural Networks and Learning Systems, abs/2110.01889:1–21, 2022.
[9] Ethan Caballero, Kshitij Gupta, Irina Rish, and David Krueger. Broken neural scaling laws. In
ICLR, 2023.
[10] Chun-Hao Chang, Rich Caruana, and Anna Goldenberg. NODE-GAM: neural generalized
additive model for interpretable deep learning. In ICLR, 2022.
[11] Jintai Chen, Kuanlun Liao, Yao Wan, Danny Z. Chen, and Jian Wu. Danets: Deep abstract
networks for tabular data classification and regression. In AAAI, pages 3930–3938, 2022.
[12] Jintai Chen, KuanLun Liao, Yanwen Fang, Danny Chen, and Jian Wu. Tabcaps: A capsule
neural network for tabular data classification with bow routing. In ICLR, 2023.
[13] Tianqi Chen and Carlos Guestrin. Xgboost: A scalable tree boosting system. In KDD, pages
785–794, 2016.
[14] Heng-Tze Cheng, Levent Koc, Jeremiah Harmsen, Tal Shaked, Tushar Chandra, Hrishi Aradhye,
Glen Anderson, Greg Corrado, Wei Chai, Mustafa Ispir, Rohan Anil, Zakaria Haque, Lichan
Hong, Vihan Jain, Xiaobing Liu, and Hemal Shah. Wide & deep learning for recommender
systems. In DLRS, pages 7–10, 2016.
[15] Corinna Cortes, Lawrence D. Jackel, Sara A. Solla, Vladimir Vapnik, and John S. Denker.
Learning curves: Asymptotic values and rate of convergence. In NIPS, pages 327–334, 1993.
[16] Manuel Fernández Delgado, Eva Cernadas, Senén Barro, and Dinani Gomes Amorim. Do we
need hundreds of classifiers to solve real world classification problems? Journal of Machine
Learning Research, 15(1):3133–3181, 2014.
[17] Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. BERT: pre-training of
deep bidirectional transformers for language understanding. In NAACL-HLT, pages 4171–4186,
2019.
[18] Yury Gorishniy, Ivan Rubachev, Valentin Khrulkov, and Artem Babenko. Revisiting deep
learning models for tabular data. In NeurIPS, pages 18932–18943, 2021.
[19] Yury Gorishniy, Ivan Rubachev, and Artem Babenko. On embeddings for numerical features in
tabular deep learning. In NeurIPS, pages 24991–25004, 2022.
[20] Yury Gorishniy, Ivan Rubachev, Nikolay Kartashev, Daniil Shlenskii, Akim Kotelnikov, and
Artem Babenko. Tabr: Tabular deep learning meets nearest neighbors in 2023. In ICLR, 2024.
[21] Léo Grinsztajn, Edouard Oyallon, and Gaël Varoquaux. Why do tree-based models still
outperform deep learning on typical tabular data? In NeurIPS, pages 507–520, 2022.
[22] Huifeng Guo, Ruiming Tang, Yunming Ye, Zhenguo Li, and Xiuqiang He. Deepfm: A
factorization-machine based neural network for CTR prediction. In IJCAI, pages 1725–1731,
2017.
10
[23] Kam Hamidieh. Superconductivty Data. UCI Machine Learning Repository, 2018. DOI:
https://doi.org/10.24432/C53P47.
[24] Md. Rafiul Hassan, Sadiq Al-Insaif, Muhammad Imtiaz Hossain, and Joarder Kamruzzaman.
A machine learning approach for prediction of pregnancy outcome following IVF treatment.
Neural Computing and Applications, 32(7):2283–2297, 2020.
[25] Joel Hestness, Sharan Narang, Newsha Ardalani, Gregory F. Diamos, Heewoo Jun, Hassan
Kianinejad, Md. Mostofa Ali Patwary, Yang Yang, and Yanqi Zhou. Deep learning scaling is
predictable, empirically. CoRR, abs/1712.00409, 2017.
[26] Derek Hoiem, Tanmay Gupta, Zhizhong Li, and Michal Shlapentokh-Rothman. Learning curves
for analysis of deep networks. In ICML, pages 4287–4296, 2021.
[27] Noah Hollmann, Samuel Müller, Katharina Eggensperger, and Frank Hutter. Tabpfn: A
transformer that solves small tabular classification problems in a second. In ICLR, 2023.
[28] Xin Huang, Ashish Khetan, Milan Cvitkovic, and Zohar S. Karnin. Tabtransformer: Tabular
data modeling using contextual embeddings. CoRR, abs/2012.06678, 2020.
[29] Alan Jeffares, Tennison Liu, Jonathan Crabbé, Fergus Imrie, and Mihaela van der Schaar. Tangos:
Regularizing tabular neural networks through gradient orthogonalization and specialization. In
ICLR, 2023.
[30] Yu-Chin Juan, Yong Zhuang, Wei-Sheng Chin, and Chih-Jen Lin. Field-aware factorization
machines for CTR prediction. In RecSys, pages 43–50, 2016.
[31] Arlind Kadra, Marius Lindauer, Frank Hutter, and Josif Grabocka. Well-tuned simple nets excel
on tabular datasets. In NeurIPS, pages 23928–23941, 2021.
[32] Jared Kaplan, Sam McCandlish, Tom Henighan, Tom B. Brown, Benjamin Chess, Rewon Child,
Scott Gray, Alec Radford, Jeffrey Wu, and Dario Amodei. Scaling laws for neural language
models. CoRR, abs/2001.08361, 2020.
[33] Guolin Ke, Qi Meng, Thomas Finley, Taifeng Wang, Wei Chen, Weidong Ma, Qiwei Ye, and
Tie-Yan Liu. Lightgbm: A highly efficient gradient boosting decision tree. In NIPS, pages
3146–3154, 2017.
[34] Günter Klambauer, Thomas Unterthiner, Andreas Mayr, and Sepp Hochreiter. Self-normalizing
neural networks. In NIPS, pages 971–980, 2017.
[35] Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. In ICLR, 2019.
[36] Duncan C. McElfresh, Sujay Khandagale, Jonathan Valverde, Vishak Prasad C., Ganesh Ra-
makrishnan, Micah Goldblum, and Colin White. When do neural nets outperform boosted trees
on tabular data? In NeurIPS, pages 76336–76369, 2023.
[37] Mehryar Mohri, Afshin Rostamizadeh, and Ameet Talwalkar. Foundations of Machine Learning.
MIT Press, 2012.
[38] Youssef Nader, Leon Sixt, and Tim Landgraf. DNNR: differential nearest neighbors regression.
In ICML, pages 16296–16317, 2022.
[39] Lennart J Nederstigt, Steven S Aanen, Damir Vandic, and Flavius Frasincar. Floppies: a
framework for large-scale ontology population of product information from tabular data in
e-commerce stores. Decision Support Systems, 59:296–311, 2014.
[40] Sergei Popov, Stanislav Morozov, and Artem Babenko. Neural oblivious decision ensembles
for deep learning on tabular data. In ICLR, 2020.
[41] Liudmila Ostroumova Prokhorenkova, Gleb Gusev, Aleksandr Vorobev, Anna Veronika Doro-
gush, and Andrey Gulin. Catboost: unbiased boosting with categorical features. In NeurIPS,
pages 6639–6649, 2018.
11
[42] Jonathan S. Rosenfeld, Amir Rosenfeld, Yonatan Belinkov, and Nir Shavit. A constructive
prediction of the generalization error across scales. In ICLR, 2020.
[43] John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. Proximal
policy optimization algorithms. CoRR, abs/1707.06347, 2017.
[44] Lisa M Schwartz, Steven Woloshin, and H Gilbert Welch. The drug facts box: providing
consumers with simple tabular data on drug benefit and harm. Medical Decision Making, 27(5):
655–662, 2007.
[45] Junhong Shen, Liam Li, Lucio M Dery, Corey Staten, Mikhail Khodak, Graham Neubig, and
Ameet Talwalkar. Cross-modal fine-tuning: Align then refine. In ICML, pages 31030–31056,
2023.
[46] Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for large-scale
image recognition. In ICLR, 2015.
[47] Weiping Song, Chence Shi, Zhiping Xiao, Zhijian Duan, Yewen Xu, Ming Zhang, and Jian
Tang. Autoint: Automatic feature interaction learning via self-attentive neural networks. In
CIKM, pages 1161–1170, 2019.
[48] Abdulhamit Subasi. Medical decision support system for diagnosis of neuromuscular disorders
using dwt and fuzzy support vector machines. Computers in Biology and Medicine, 42(8):
806–815, 2012.
[49] Joaquin Vanschoren, Jan N Van Rijn, Bernd Bischl, and Luis Torgo. Openml: networked science
in machine learning. ACM SIGKDD Explorations Newsletter, 15(2):49–60, 2014.
[50] Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez,
Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. In NIPS, 2017.
[51] Oriol Vinyals, Charles Blundell, Tim Lillicrap, Koray Kavukcuoglu, and Daan Wierstra. Match-
ing networks for one shot learning. In NIPS, pages 3630–3638, 2016.
[52] Pauli Virtanen, Ralf Gommers, Travis E. Oliphant, Matt Haberland, Tyler Reddy, David
Cournapeau, Evgeni Burovski, Pearu Peterson, Warren Weckesser, Jonathan Bright, Stéfan J.
van der Walt, Matthew Brett, Joshua Wilson, K. Jarrod Millman, Nikolay Mayorov, Andrew
R. J. Nelson, Eric Jones, Robert Kern, Eric Larson, C J Carey, İlhan Polat, Yu Feng, Eric W.
Moore, Jake VanderPlas, Denis Laxalde, Josef Perktold, Robert Cimrman, Ian Henriksen, E. A.
Quintero, Charles R. Harris, Anne M. Archibald, Antônio H. Ribeiro, Fabian Pedregosa, Paul
van Mulbregt, and SciPy 1.0 Contributors. SciPy 1.0: Fundamental Algorithms for Scientific
Computing in Python. Nature Methods, 17:261–272, 2020.
[53] Ruoxi Wang, Rakesh Shivanna, Derek Zhiyuan Cheng, Sagar Jain, Dong Lin, Lichan Hong,
and Ed H. Chi. DCN V2: improved deep & cross network and practical lessons for web-scale
learning to rank systems. In WWW, pages 1785–1797, 2021.
[54] Jing Wu, Suiyao Chen, Qi Zhao, Renat Sergazinov, Chen Li, Shengjie Liu, Chongchao Zhao,
Tianpei Xie, Hanqing Guo, Cheng Ji, Daniel Cociorva, and Hakan Brunzell. Switchtab:
Switched autoencoders are effective tabular learners. In AAAI, pages 15924–15933, 2024.
[55] Ling Yan, Wu-Jun Li, Gui-Rong Xue, and Dingyi Han. Coupled group lasso for web-scale CTR
prediction in display advertising. In ICML, pages 802–810, 2014.
[56] Han-Jia Ye, De-Chuan Zhan, Xue-Min Si, and Yuan Jiang. Learning mahalanobis distance
metric: Considering instance disturbance helps. In IJCAI, pages 3315–3321, 2017.
[57] Han-Jia Ye, De-Chuan Zhan, Nan Li, and Yuan Jiang. Learning multiple local metrics: Global
consideration helps. IEEE Trans. Pattern Anal. Mach. Intell., 42(7):1698–1712, 2020.
[58] Han-Jia Ye, Qi-Le Zhou, and De-Chuan Zhan. Training-free generalization on heterogeneous
tabular data via meta-representation. CoRR, abs/2311.00055, 2023.
[59] Hangting Ye, Wei Fan, Xiaozhuang Song, Shun Zheng, He Zhao, Dan dan Guo, and Yi Chang.
Ptarl: Prototype-based tabular representation learning via space calibration. In ICLR, 2024.
12
[60] Weinan Zhang, Tianming Du, and Jun Wang. Deep learning over multi-field categorical data - -
A case study on user response prediction. In ECIR, pages 45–57, 2016.
[61] Qi-Le Zhou, Han-Jia Ye, Leye Wang, and De-Chuan Zhan. Unlocking the transferability of
tokens in deep models for tabular data. CoRR, abs/2310.15149, 2023.
[62] Bingzhao Zhu, Xingjian Shi, Nick Erickson, Mu Li, George Karypis, and Mahsa Shoaran. Xtab:
Cross-table pretraining for tabular transformers. In ICML, pages 43181–43204, 2023.
13
Appendix A Benchmark Details
As mentioned in section 3 in the main paper, we propose a large benchmark for the comprehensive
evaluation of tabular methods, which contains 101 binary classification datasets, 80 multi-class
classification datasets, and 119 regression datasets. In this section, we present detailed descriptions
of the 300 datasets, as well as their statistics.
In Table 4, we list the names as well as the source URLs of all datasets in our benchmark. We select
these datasets based on the rules mentioned in section 3. Each dataset is associated with an ID number.
In the following tables, we use the ID to index a dataset and avoid listing long dataset names.
Table 4: The list of datasets (names as well as source URLs) in our proposed benchmark.
14
67 INNHotelsGroup 217 mfeat-pixel
68 Indian_pines 218 mfeat-zernike
69 Insurance 219 mice_protein_expression
70 Intersectional-Bias-Assessment 220 microaggregation2
71 Is-this-a-good-customer 221 mobile_c36_oversampling
72 JapaneseVowels 222 mozilla4
73 Job_Profitability 223 mv
74 KDD 224 naticusdroid+android+permissions+dataset
75 KDDCup09_upselling 225 national-longitudinal-survey-binary
76 Kaggle_bike_sharing_demand_challange 226 okcupid_stem
77 Laptop_Prices_Dataset 227 one-hundred-plants-margin
78 Large-scale_Wave_Energy_Farm_Perth_100 228 one-hundred-plants-shape
79 Large-scale_Wave_Energy_Farm_Perth_49 229 one-hundred-plants-texture
80 Large-scale_Wave_Energy_Farm_Sydney_100 230 online_shoppers
81 Large-scale_Wave_Energy_Farm_Sydney_49 231 optdigits
82 Long 232 ozone-level-8hr
83 MIC 233 ozone_level
84 MagicTelescope 234 page-blocks
85 Marketing_Campaign 235 pc1
86 MiamiHousing2016 236 pc3
87 Mobile_Phone_Market_in_Ghana 237 pc4
88 Mobile_Price_Classification 238 pendigits
89 NASA_PHM2008 239 phoneme
90 NHANES_age_prediction 240 pol
91 National_Health_and_Nutrition_Health_Survey 241 pol_reg
92 Parkinson_Multiple_Sound_Recording 242 pole
93 Parkinsons_Telemonitoring 243 predict_students_dropout_and_academic_success
94 Performance-Prediction 244 puma32H
95 PhishingWebsites 245 puma8NH
96 Physicochemical_r 246 qsar
97 PieChart3 247 qsar_aquatic_toxicity
98 Pima_Indians_Diabetes_Database 248 qsar_fish_toxicity
99 PizzaCutter3 249 rice_cammeo_and_osmancik
100 Pumpkin_Seeds 250 ringnorm
101 QSAR_biodegradation 251 rl
102 Rain_in_Australia 252 satellite_image
103 SDSS17 253 satimage
104 Satellite 254 segment
105 Shipping 255 seismic+bumps
106 Shop_Customer_Data 256 semeion
107 Smoking_and_Drinking_Dataset_with_body_signal 257 sensory
108 Student_Alcohol_Consumption 258 shill-bidding
109 Student_Performance_Portuguese 259 shrutime
110 Superconductivty 260 shuttle
111 Telecom_Churn_Dataset 261 socmob
112 UJI_Pen_Characters 262 space_ga
113 UJIndoorLoc 263 spambase
114 VulNoneVul 264 splice
115 Water_Quality_and_Potability 265 sports_articles_for_objectivity_analysis
116 Waterstress 266 statlog
117 Wilt 267 steel_industry_energy_consumption
118 Wine_Quality_red 268 steel_plates_faults
119 Wine_Quality_white 269 stock
120 abalone 270 stock_fardamento02
121 accelerometer 271 sulfur
122 ada 272 svmguide3
123 ada_agnostic 273 sylvine
124 ada_prior 274 taiwanese_bankruptcy_prediction
125 airfoil_self_noise 275 telco-customer-churn
126 airlines_seed_0_nrows_2000_nclasses_10_ncols_100_stratify_True 276 texture
127 allbp 277 thyroid
128 allrep 278 thyroid-ann
129 analcatdata_authorship 279 thyroid-dis
130 analcatdata_supreme 280 topo_2_1
131 archive2 281 treasury
132 archive_r56_Maths 282 turiye_student_evaluation
133 archive_r56_Portuguese 283 twonorm
134 artificial-characters 284 vehicle
135 auction_verification 285 volkert
136 autoUniv-au4-2500 286 volume
137 autoUniv-au7-1100 287 walking-activity
138 avocado_sales 288 wall-robot-navigation
139 bank 289 water_quality
140 bank32nh 290 waveform-5000
141 bank8FM 291 waveform_database_generator
142 banknote_authentication 292 waveform_database_generator_version_1
143 baseball 293 weather_izmir
144 blogfeedback 294 website_phishing
145 car-evaluation 295 wind
146 churn 296 wine
147 cmc 297 wine+quality
148 combined_cycle_power_plant 298 wine-quality-red
149 communities_and_crime 299 wine-quality-white
150 company_bankruptcy_prediction 300 yeast
We list the statistics of all 300 datasets in Table 5. The datasets are indexed based on their IDs. For
each dataset, we list the type of the task (i.e., classification “CLS” or regression “REG”), the number
15
of classes (only for classification tasks), the number of samples, the number of numerical features,
the number of category features, and the domain of the dataset based on its source information (there
are 13 domains, i.e., financial, physical, biological, healthcare, natural, industrial, societal, vision,
computer, educational, leisure, handcrafted, and others). We show the statistics of datasets in Figure 2
in the main paper, which indicates the advantages of the proposed benchmark w.r.t. its coverage of
tasks, sizes, and domains.
Moreover, we also display the score of each dataset to indicate the preference of a dataset on tree-
based or DNN-based models (a higher score means the dataset is Tree-friendly and DNN-friendly
otherwise). More detailed rules for calculating Tree-DNN scores can be found in subsection B.2.
The last two columns in the table for each dataset mean whether the dataset is selected in our proposed
two tiny benchmarks.
Table 5: Statistics of all datasets. ID continues the ID in Table 4, TB1 and TB2 indicate whether this
dataset belongs to Tiny Benchmark 1 or Tiny Benchmark 2.
ID Task Classes Samples Num. Cat. Domain Score TB1 TB2 ID Task Classes Samples Num. Cat. Domain Score TB1 TB2
16
66 Reg - 4046 27 0 Computer -1.247 ✓ ✓ 216 Cls 10 2000 6 0 Physical -1.902
67 Cls 2 36275 11 6 Societal 1.337 217 Cls 10 2000 0 240 Physical -0.777
68 Cls 8 9144 220 0 Physical -1.348 218 Cls 10 2000 47 0 Physical -1.884
69 Reg - 23548 3 7 Financial 0.846 219 Cls 8 1080 75 0 Biological -0.736
70 Reg - 11000 14 5 Computer 0.020 ✓ 220 Cls 5 20000 20 0 Biological -1.440
71 Reg - 1723 9 4 Financial 0.960 221 Cls 2 51760 6 0 Others -0.638
72 Cls 9 9961 14 0 Vision -0.978 222 Cls 2 15545 4 0 Others 1.401
73 Reg - 14480 27 1 Educational 2.659 ✓ 223 Reg - 40768 7 3 Others -0.982 ✓
74 Reg - 5032 34 11 Physical 0.138 ✓ 224 Cls 2 29332 0 86 Leisure -0.249
75 Cls 2 5128 34 15 Financial 0.563 225 Cls 2 4908 9 7 Societal 0.398
76 Reg - 10886 3 6 Industrial 2.034 ✓ 226 Cls 3 26677 2 11 Societal 0.181 ✓
77 Reg - 4441 8 0 Financial 1.347 227 Cls 100 1600 64 0 Vision 1.068
78 Reg - 7277 201 0 Natural 1.146 228 Cls 100 1600 64 0 Vision -1.259
79 Reg - 36043 99 0 Natural 1.226 229 Cls 100 1599 64 0 Vision 0.162
80 Reg - 2318 201 0 Natural 0.437 230 Cls 2 12330 5 9 Financial 1.005 ✓
81 Reg - 17964 99 0 Natural -0.166 ✓ 231 Cls 10 5620 64 0 Physical -1.554 ✓
82 Reg - 4477 19 0 Others -0.159 232 Cls 2 2534 72 0 Natural 0.334
83 Cls 2 1649 9 95 Physical 1.144 233 Reg - 2536 0 36 Natural -0.235
84 Cls 2 19020 9 0 Physical 0.816 234 Cls 5 5473 10 0 Others 2.573 ✓ ✓
85 Cls 2 2240 19 8 Financial 0.879 235 Cls 2 1109 21 0 Physical 1.240
86 Reg - 13932 16 0 Financial 1.471 236 Cls 2 1563 37 0 Physical -0.680 ✓
87 Reg - 3600 4 10 Financial 0.634 237 Cls 2 1458 37 0 Physical 1.743
88 Cls 4 2000 14 6 Financial -0.037 ✓ 238 Cls 10 10992 16 0 Vision -0.193 ✓
89 Reg - 45918 21 0 Industrial -0.360 239 Cls 2 5404 5 0 Others 0.758
90 Reg - 2277 4 3 Healthcare -1.840 ✓ 240 Cls 2 10082 26 0 Others -0.933 ✓
91 Cls 2 2278 7 0 Healthcare 1.732 241 Reg - 15000 48 0 Others -1.686 ✓
92 Reg - 1040 26 0 Healthcare 1.067 242 Reg - 14998 26 0 Others -1.495 ✓
93 Reg - 5875 18 1 Healthcare 2.019 243 Cls 3 4424 5 29 Educational 1.438
94 Reg - 1340 19 0 Others 0.556 244 Reg - 8192 32 0 Industrial -0.084
95 Cls 2 11055 0 30 Leisure 0.454 245 Reg - 8192 8 0 Industrial 0.362
96 Reg - 45730 9 0 Physical -0.192 246 Cls 2 1055 30 10 Biological -2.103 ✓
97 Cls 2 1077 37 0 Others -0.351 247 Reg - 546 8 0 Biological 1.840
98 Cls 2 768 8 0 Healthcare 1.765 248 Reg - 908 4 2 Biological 1.355
99 Cls 2 1043 37 0 Financial -0.840 249 Cls 2 3810 7 0 Vision 0.851
100 Cls 2 2500 12 0 Biological 1.959 250 Cls 2 7400 20 0 Handcrafted -0.359
101 Cls 2 1054 41 0 Biological 0.608 251 Cls 2 4970 5 7 Others 1.299 ✓
102 Cls 3 145460 12 6 Natural 1.289 252 Reg - 6435 36 0 Vision 0.854
103 Cls 3 100000 8 4 Natural 1.632 253 Cls 6 6430 36 0 Vision 1.432 ✓
104 Cls 2 5100 36 0 Vision 0.197 254 Cls 7 2310 17 0 Vision 0.112 ✓
105 Reg - 10999 5 4 Industrial 0.273 255 Cls 2 2584 14 4 Natural 1.593
106 Reg - 2000 4 2 Financial 0.149 ✓ 256 Cls 10 1593 256 0 Physical -1.288
107 Cls 2 991346 22 1 Healthcare 0.180 257 Reg - 576 0 11 Vision 0.456
108 Reg - 395 13 17 Educational 2.047 ✓ ✓ 258 Reg - 6321 3 0 Leisure -0.134
109 Reg - 397 15 17 Educational 0.793 259 Reg - 10000 4 6 Others 0.715
110 Reg - 21197 81 0 Industrial 1.815 ✓ 260 Cls 7 58000 9 0 Physical 0.751
111 Cls 2 3333 14 3 Financial 0.250 261 Reg - 1156 1 4 Others -2.069 ✓
112 Cls 35 1364 80 0 Physical -2.218 262 Reg - 3107 6 0 Others -1.406
113 Reg - 21048 520 0 Physical 0.172 263 Cls 2 4601 57 0 Physical 2.083
114 Reg - 5692 16 0 Industrial 0.032 ✓ ✓ 264 Cls 3 3190 0 60 Physical 0.173 ✓
115 Cls 2 3276 8 0 Natural -0.306 265 Cls 2 1000 57 2 Leisure 1.452
116 Reg - 1188 22 0 Natural -0.045 ✓ 266 Cls 2 1000 7 13 Others 1.435
117 Cls 2 4821 5 0 Natural -0.100 267 Reg - 35040 7 3 Industrial -0.920
118 Reg - 1599 11 0 Vision 1.820 268 Cls 7 1941 25 2 Industrial 1.904
119 Reg - 4898 11 0 Vision 1.577 269 Reg - 950 9 0 Financial -0.789
120 Cls 3 4177 7 1 Biological -1.728 270 Reg - 6277 5 1 Financial 0.596
121 Cls 4 153004 3 1 Physical -1.214 271 Reg - 10081 6 0 Natural -1.680
122 Cls 2 4147 48 0 Physical 1.718 272 Reg - 1243 22 0 Others 1.799
123 Cls 2 4562 48 0 Physical 1.477 273 Cls 2 5124 20 0 Physical 0.618 ✓
124 Cls 2 4562 6 8 Physical 0.301 274 Cls 2 6819 95 0 Societal 1.686 ✓
125 Reg - 1503 5 0 Industrial -0.474 275 Cls 2 7043 3 15 Financial 1.002
126 Cls 2 2000 3 4 Industrial 1.151 276 Cls 11 5500 40 0 Vision -0.679
127 Cls 3 3772 6 23 Biological 1.683 277 Cls 3 7200 21 0 Healthcare 1.582
128 Cls 4 3772 6 23 Biological 1.555 278 Cls 3 3772 21 0 Healthcare 1.829
129 Cls 4 841 69 0 Educational 0.135 279 Cls 5 2800 6 20 Healthcare -2.415 ✓
130 Reg - 4052 7 0 Educational 1.949 280 Reg - 8885 266 0 Natural 1.299
131 Reg - 1143 11 1 Educational 1.194 ✓ 281 Reg - 1049 15 0 Financial -1.689
132 Reg - 397 1 29 Educational -0.332 282 Cls 5 5820 30 2 Educational 0.250
133 Reg - 651 1 29 Educational 2.352 ✓ 283 Cls 2 7400 20 0 Handcrafted -0.542
134 Cls 10 10218 7 0 Physical 1.633 284 Cls 4 846 18 0 Industrial -2.127
135 Reg - 2043 6 1 Financial 1.957 285 Cls 10 58310 180 0 Others -1.075
136 Cls 3 2500 58 42 Handcrafted 0.662 286 Reg - 50993 53 0 Others 0.632
137 Cls 5 1100 8 4 Handcrafted 0.234 287 Cls 22 149332 4 0 Industrial 0.559
138 Reg - 18249 11 2 Financial 0.237 288 Cls 4 5456 24 0 Industrial 0.635
139 Cls 2 45211 7 9 Financial -0.496 289 Cls 2 7996 20 0 Natural -1.962
140 Reg - 8192 32 0 Financial -0.187 290 Cls 3 5000 40 0 Physical -0.666 ✓
141 Reg - 8192 8 0 Financial -0.131 ✓ 291 Cls 3 4999 21 0 Physical 1.676
142 Cls 2 1372 4 0 Financial -1.064 292 Cls 3 5000 21 0 Physical -1.498
143 Cls 3 1340 15 1 Leisure 0.737 ✓ 293 Reg - 1461 9 0 Natural -0.043
144 Reg - 60021 62 214 Physical 1.687 294 Cls 3 1353 0 9 Computer 0.322 ✓
145 Cls 4 1728 0 21 Physical 0.331 295 Reg - 6574 14 0 Natural -0.577
146 Cls 2 5000 16 4 Financial -0.567 296 Cls 2 2554 4 0 Vision 2.235 ✓
147 Cls 3 1473 2 7 Physical 1.697 297 Reg - 6497 11 0 Vision 1.747
148 Reg - 9568 4 0 Industrial 1.922 298 Cls 6 1599 4 0 Vision 2.143
149 Reg - 1994 102 0 Societal 0.401 ✓ 299 Cls 7 4898 11 0 Vision 1.756 ✓
150 Cls 2 6819 93 2 Financial 1.086 300 Cls 10 1484 8 0 Biological 2.365 ✓
Appendix B Discussions
We discuss additional details and setups to construct our proposed (tiny) benchmarks.
17
B.1 Additional Details to Construct the Whole Benchmark
The search space of hyper-parameters. As mentioned in subsection 3.1 in the main paper, all hyper-
parameters are searched using Optuna [2] over 100 trials. We list the range of hyper-parameters for
each comparison method below for reproduction. For most methods, we follow the hyper-parameter
space provided in [18]. For more details, see our source code.
• MLP
– layers: UniformInt[1, 8]
– d_layers: UniformInt[64, 512]
– dropout: {0, Uniform[0, 0.5]}
– lr: LogUniform[1e-5, 1e-2]
– weight_decay: {0, LogUniform[1e-6, 1e-3]}
• ResNet
– layers: UniformInt[1, 8]
– d_layers: UniformInt[64, 512]
– dropout: {0, Uniform[0, 0.5]}
– lr: LogUniform[1e-5, 1e-2]
– weight_decay: {0, LogUniform[1e-6, 1e-3]}
• FT-T
– n_layers: UniformInt[1, 4]
– d_token: Categorical{8, 16, 32, 64, 128}
– residual_dropout: {0, Uniform[0, 0.2]}
– attention_dropout: Uniform[0, 0.5]
– ffn_dropout: Uniform[0, 0.5]
– d_ffn_factor: Uniform[0.667, 2.667]
– lr: LogUniform[1e-5, 1e-3]
– weight_decay: LogUniform[1e-6, 1e-3]
• NODE
– num_layers: UniformInt[1, 4]
– depth: UniformInt[4, 6]
– tree_dim: UniformInt[2, 3]
– layer_dim: Categorical{512, 1024}
– lr: LogUniform[1e-5, 0.1]
– weight_decay: LogUniform[1e-6, 1e-3]
• TabR
– d_main: UniformInt[96, 384]
– context_dropout: Uniform[0, 0.6]
– encoder_n_blocks: UniformInt[0, 1]
– predictor_n_blocks: UniformInt[1, 2]
– dropout0: Uniform[0, 0.6]
– num_embeddings:
* n_frequencies: UniformInt[16, 96]
* frequency_scale: LogUniform[0.01, 100.0]
* d_embedding: UniformInt[16, 64]
– lr: LogUniform[3e-5, 1e-3]
– weight_decay: {0, LogUniform[1e-6, 1e-3]}
• TANGOS
– layers: UniformInt[1, 8]
– d_layers: UniformInt[64, 512]
– dropout: {0, Uniform[0, 0.5]}
– lambda1: LogUniform[0.001, 10]
– lambda2: LogUniform[0.0001, 1]
– subsample: UniformInt[30, 100]
– lr: LogUniform[1e-4, 1e-3]
– weight_decay: {0, LogUniform[1e-6, 1e-3]}
• TabCaps
– lr: LogUniform[1e-5, 0.1]
– weight_decay: LogUniform[1e-6, 1e-3]
– sub_class: UniformInt[1, 5]
18
– init_dim: UniformInt[32, 128]
– primary_capsule_size: UniformInt[4, 32]
– digit_capsule_size: UniformInt[4, 32]
– leaves: UniformInt[16, 64]
• TabNet
– lr: Uniform[0.001, 0.01]
– gamma: Uniform[1, 2]
– n_steps: UniformInt[3, 10]
– n_independent: UniformInt[1, 5]
– n_shared: UniformInt[1, 5]
– momentum: Uniform[0.01, 0.4]
• SNN
– layers: UniformInt[2, 16]
– d_layers: UniformInt[1, 512]
– dropout: {0, Uniform[0, 0.1]}
– d_embedding: UniformInt[64, 512]
– lr: LogUniform[1e-5, 1e-2]
– weight_decay: {0, LogUniform[1e-6, 1e-3]}
• PTaRL
– layers: UniformInt[1, 3]
– d_layers: UniformInt[64, 512]
– dropout: {0, Uniform[0, 0.5]}
– d_embedding: UniformInt[64, 128]
– lr: LogUniform[1e-5, 0.1]
– weight_decay: {0, LogUniform[1e-6, 1e-3]}
• DANets
– n_layers: UniformInt[6, 32]
– dropout: {0, Uniform[0, 0.5]}
– base_outdim: UniformInt[64, 128]
– lr: LogUniform[1e-5, 0.1]
– weight_decay: {0, LogUniform[1e-6, 1e-3]}
• DCNv2
– d: UniformInt[64, 512]
– d_embedding: UniformInt[64, 512]
– hidden_dropout: Uniform[0, 0.5]
– cross_dropout: Uniform[0, 0.5]
– n_cross_layers: UniformInt[1, 8]
– n_hidden_layers: UniformInt[1, 8]
– lr: LogUniform[1e-5, 0.1]
– weight_decay: {0, LogUniform[1e-6, 1e-3]}
• TabTransformer
– dim: Categorical{32, 64, 128, 256}
– depth: Categorical{1, 2, 3, 6, 12}
– heads: Categorical{2, 4, 8}
– attn_dropout: Uniform[0, 0.5]
– ff_dropout: Uniform[0, 0.5]
– lr: LogUniform[1e-5, 0.1]
– weight_decay: {0, LogUniform[1e-6, 1e-3]}
• GrowNet
– d_embedding: UniformInt[32, 512]
– hidden_d: UniformInt[32, 512]
– lr: LogUniform[1e-5, 0.1]
– weight_decay: {0, LogUniform[1e-6, 1e-3]}
– epochs_per_stage: UniformInt[1, 2]
– correct_epoch: UniformInt[1, 2]
• AutoInt
– n_layers: UniformInt[1, 6]
– d_token: Categorical{8, 16, 32, 64, 128}
– residual_dropout: {0, Uniform[0, 0.2]}
19
– attention_dropout: Uniform[0, 0.5]
– lr: LogUniform[1e-5, 1e-3]
– weight_decay: LogUniform[1e-6, 1e-3]
• DNNR
– n_neighbors: Fixed[3]
– n_derivative_neighbors: UniformInt[32, 1024]
– order: Categorical{"1", "2diag"}
– solver: Categorical{"linear_regression", "scipy_lsqr", "numpy", "ridge", "lasso"}
– index: Fixed{"annoy"}
• SwitchTab
– alpha: LogUniform[0.01, 100]
– lr: LogUniform[1e-6, 1e-3]
– weight_decay: LogUniform[1e-6, 1e-3]
• XGBoost
– alpha: {0, LogUniform[1e-8, 100]}
– colsample_bylevel: Uniform[0.5, 1.0]
– colsample_bytree: Uniform[0.5, 1.0]
– gamma: {0, LogUniform[1e-8, 100]}
– lambda: {0, LogUniform[1e-8, 100]}
– learning_rate: LogUniform[1e-5, 1]
– max_depth: UniformInt[3, 10]
– min_child_weight: LogUniform[1e-8, 100000]
– subsample: Uniform[0.5, 1.0]
• CatBoost
– bagging_temperature: Uniform[0, 1.0]
– depth: UniformInt[3, 10]
– l2_leaf_reg: LogUniform[1.0, 10.0]
– leaf_estimation_iterations: UniformInt[1, 10]
– learning_rate: LogUniform[1e-5, 1]
• LightGBM
– num_leaves: UniformInt[10, 100]
– max_depth: UniformInt[3, 10]
– learning_rate: LogUniform[1e-3, 1]
– min_child_weight: LogUniform[1e-5, 1e-1]
– min_child_samples: UniformInt[2, 100]
– subsample: Uniform[0.5, 1.0]
– colsample_bytree: Uniform[0.5, 1.0]
– reg_lambda: {0, LogUniform[1e-5, 1.0]}
• LogReg
– C: LogUniform[1e-5, 5]
– penalty: Categorical{"l2", null}
– max_iter: UniformInt[50, 500]
• Random Forest
– min_samples_split: UniformInt[2, 10]
– min_samples_leaf: UniformInt[1, 10]
• SVM
– C: LogUniform[1e-5, 1e5]
• KNN
– n_neighbors: UniformInt[1, 128]
– weights: Categorical{"uniform", "distance"}
– p: Categorical{1, 1.5, 2, 2.5, 3}
The content given a dataset and method pair. Our benchmark provides a comprehensive evaluation
of various tabular methods across numerous datasets. For each dataset-method pair, we record
extensive statistics, which are valuable in tabular machine learning research. The recorded statistics
include:
• Training logs. Detailed logs are kept for all methods, capturing changes in statistics such as
learning rate, batch-wise losses, and other criteria throughout the training process.
20
• Performance metrics. For each method, hyper-parameters are tuned over 100 trials using the
training and validation sets. The best-performing hyper-parameters are then used to train the model
with 15 random seeds, and the average performance is reported. We apply early stopping based on
accuracy for classification tasks and RMSE for regression tasks. In addition to tracking validation
and test set statistics (loss values, accuracy, and RMSE) throughout the training epochs, we also
record test set criteria. Additional metrics such as F1/AUC for classification and MAE/R2 for
regression are used to evaluate test set performance.
• Running time. The running time for each method is recorded over the 15 random seeds. Since
models may stop early at different epochs, running times can vary across epochs.
• Model sizes. The sizes of models with default hyper-parameters and those with tuned hyper-
parameters are recorded.
We construct two tiny benchmarks with 15% of the whole size to facilitate further tabular research
in a lightweight manner. In particular, Tiny Benchmark 1 is for Tree/DNN comparison, and Tiny
Benchmark 2 is for rank-consistent evaluation. We will illustrate the selection method respectively.
Tiny Benchmark 1 for Tree/DNN comparison. We expect this tiny benchmark to cover datasets in
the whole benchmark with different sizes (N × d) and have diverse preferences for tree-based and
DNN-based methods.
To differentiate the tree-friendly and DNN-friendly datasets, we introduce a Tree-DNN score as
displayed in Table 5, which indicates the gap between tree-based methods and DNN-based methods.
We first choose 6 representative tabular methods: XGBoost, CatBoost, Random Forest, MLP, ResNet,
and FT-T. The first three are tree-based methods, and the remaining ones represent DNN-based
methods. For each dataset, the results of the selected methods are first normalized to remove the
differences in the difficulty level of each dataset. We then use the difference between the best-
performed tree-based methods and the best-performing DNN-based method as a score reflecting the
gap between them. Formally, the Tree-DNN score is given by:
s = max(ŝXGBoost , ŝCatBoost , ŝRForest ) − max(ŝMLP , ŝResNet , ŝFT-T ) , (5)
where ŝ is the normalized metric, i.e., the accuracy for the classification task, or negative RMSE for
the regression task.
We then extract 15% of the dataset by task type, i.e., 15 binary classification tasks, 12 multi-class
classification tasks, and 18 regression tasks are selected from the 300 datasets. More specifically, we
sort the datasets for each task type by their size (N × d) and divide the 101 binary classification tasks,
80 multi-class classification tasks, and 119 regression tasks into 5, 4, and 6 groups, respectively. For
each group, we choose one dataset from each of the three cases: “TF”, “DF”, and “Tie”, which means
choosing the dataset with the highest score, the lowest score, and the score with an absolute value
close to zero. We also adjust the selection by choosing a second candidate dataset, in order to ensure
that the number of datasets containing categorical features and datasets containing numerical features
in the tiny benchmark are balanced. Such a selection method helps us to select datasets with different
performances while taking into account the range of dataset size and the distribution of features.
Tiny Benchmark 2 for rank consistent evaluation. We also extract 15% of the datasets by task
type to form Tiny Benchmark 2, aiming to maintain the rank consistency with the full benchmark,
rather than its diversity. To minimize Equation 4 in the main paper, we consider three approximated
approaches: greedy selection, random selection, and clustering-based selection. The criterion is the
mean absolute error (MAE) between the ranks of methods on the selected subsets and the ranks of
methods on the whole dataset. The details of the three strategies are:
• Greedy Selection. For each task type, we specify the number of datasets to choose, as mentioned
above. Then, in each loop, we select the dataset that minimizes the MAE until the desired number
is reached.
• Random Selection. For each task type, we randomly select a specified number of datasets for
10,000 trials. Then, we choose the subsets with the minimum MAE.
• Clustering. The goal of selecting a subset of datasets such that the ranks of methods on it
approximate their ranks on the full benchmark can be viewed as a clustering problem. We perform
KMeans clustering with Euclidean distance and select the dataset closest to each cluster center as
the representative.
21
We first test the performance of the above three approaches by performing multiple splits on these
13 methods: Dummy, LR, KNN, SVM, XGBoost, CatBoost, Random Forest, MLP, ResNet, FT-T,
DCNv2, TabR, and TabPFN. In each split, we choose 10 methods as seen methods and others as
unseen methods (i.e., 3 for classification tasks and 2 for regression tasks since TabPFN cannot be
applied to regression tasks). The results are shown in Table 2 in the main paper.
We find that the greedy selection does not work well, while random selection achieves the best
performance on both seen and unseen methods. The possible reason is that since we specify the
number of datasets to choose, the greedy selection might not achieve the optimum within that number.
The best performance was observed in the randomly selected set, which, to some extent, also indicates
that our dataset is distributed quite uniformly, rather than conforming to the assumptions of clustering.
Therefore, we release Tiny Benchmark 2 by performing random selection on all 13 methods as
mentioned above, as shown in Table 5.
In section 4 of the main paper, we introduce a novel tabular prediction setting where we focus on
predicting the training dynamics of a model. Specifically, using the meta-features of a dataset, we aim
to predict the validation curves of a model as well as its overall training progress. The meta-features
offer valuable information about the dataset, enhancing the accuracy of our model predictions. All the
meta-features utilized in this prediction task are detailed in Table 6, accompanied by their respective
explanations. These meta-features play a crucial role in understanding the dataset characteristics and
are instrumental in accurately forecasting the model’s training behavior.
In this section, we describe the details of training a neural network for curve prediction and compare
it with conventional scaling law based methods, such as M1, M2 [15, 25, 42, 1], M3 [43], and M4 [3].
We also provide specific evaluation metrics on the task.
22
Table 6: Meta-features used in the training dynamics prediction task. The first column indicates the
selected meta-feature by the model in the training dynamics prediction task.
Selected Meta-Feature Explanation
attr_conc The concentration coef. of each pair of distinct attributes.
✓ class_conc The concentration coefficient between each attribute and class.
class_ent The target attribute Shannon’s entropy.
✓ inst_to_attr The ratio between the number of instances and attributes.
✓ mean The mean value of each attribute.
sd The standard deviation of each attribute.
var The variance of each attribute.
✓ range The range (max - min) of each attribute.
✓ iq_range The interquartile range (IQR) of each attribute.
✓ nr_attr The total number of attributes.
✓ sparsity The (possibly normalized) sparsity metric for each attribute.
t_mean The trimmed mean of each attribute.
nr_bin The number of binary attributes.
nr_cat The number of categorical attributes.
nr_num The number of numeric features.
nr_norm The number of attributes normally distributed based in a given method.
nr_cor_attr The number of distinct highly correlated pair of attributes.
✓ gravity The distance between minority and majority classes’ center of mass.
nr_class The number of distinct classes.
✓ joint_ent The joint entropy between each attribute and class.
✓ attr_ent Shannon’s entropy for each predictive attribute.
✓ cov The absolute value of the covariance of distinct dataset attribute pairs.
eigenvalues The eigenvalues of covariance matrix from dataset.
eq_num_attr The number of attributes equivalent for a predictive task.
✓ max The maximum value from each attribute.
min The minimum value from each attribute.
median The median value from each attribute.
freq_class The relative frequency of each distinct class.
mad The Median Absolute Deviation (MAD) adjusted by a factor.
mad The Median Absolute Deviation (MAD) adjusted by a factor.
✓ mut_inf The mutual information between each attribute and target.
✓ nr_inst The number of instances (rows) in the dataset.
✓ nr_outliers The number of attributes with at least one outlier value.
✓ ns_ratio The noisiness of attributes.
✓ imblance_ratio The ratio of the number of instances in the minority to the majority class.
attr_to_inst The ratio between the number of attributes.
features and their derived metrics, such as range.mean, and range.std, listed in Table 6), and its output
dimension is 4, corresponding to the parameters in θ.
To learn the mapping h, we minimize the MAE between the predicted curves and the actual curves
on the training set of the task. In detail, we input the first 5 epochs’ validation performance and
the meta-features of the corresponding dataset into the MLP to obtain the predicted scaling law
parameters. These parameters are then substituted into Equation 6 to generate the predicted curves.
We calculate the MAE between the predicted and actual curves and back-propagate the loss to adjust
the MLP’s weights. After several iterations of training, we obtain a curve predictor that can accurately
forecast the validation curves.
We compare our method with five different scaling laws, whose parameters could be estimated based
on the first five points of the validation curve.
• M1 describes the basic form of the scaling law: y = axb . This has been used, for example, to
estimate the required sample size in language models [32].
• M2 [15, 25, 42, 1] models the performance of multi-layer neural networks as a power law of the
form y = axb + c.
23
• M3 [43] introduces the functional form y = a(x + d)b + c, where d represents the scale at which
the performance starts to improve beyond the random guess loss and transitions to a power law
scaling regime.
• M4 [3] proposes the functional form (y − ϵ∞ )/((ϵ0 − y)a ) = bxc , where ϵ∞ is the irreducible
entropy of the data distribution and ϵ0 is the random guess
√ performance.
• Our Scaling Law represents the form y = A log x + B x + C + D/x, without training an MLP.
It should be noted that we used the official code to fit the scaling law parameters mentioned above,
available at revisiting_neural_scaling_laws.
We list the detailed results of the Performance-Efficiency-Size comparison in Figure 1 in the main
paper in Table 8, which contains the average rank, running time, and checkpoint size of each method.
24
Table 7: Average MAE and OVD for various curves of test datasets.
Comparison Methods MAE OVD
M1 0.1858 0.4575
M2 0.8383 8.7971
M3 0.1351 0.1577
M4 0.1851 0.1849
Our Scaling law 1.1949 1.9854
Ours with MLP 0.0928 0.1164
1.0
0.75
1.0
0.9 1.0
0.70 0.9
0.8
Valid Accuracy
Valid Accuracy
Valid Accuracy
Valid Accuracy
0.9
0.8
0.65
0.7
our sacling law our sacling law 0.8
our sacling law our sacling law
0.60 M1 0.6 M1 M1 0.7
M1
M2 M2 0.7 M2 M2
M3 0.5 M3 M3 0.6
M3
0.55
M4 M4 0.6
M4 M4
Ours with MLP 0.4 Ours with MLP Ours with MLP 0.5
Ours with MLP
0.50 Train Data Train Data Train Data Train Data
Test Data 0.3 Test Data 0.5 Test Data 0.4 Test Data
Epoch Epoch Epoch Epoch
0 10 20 30 40 50 60 70 0 25 50 75 100 125 150 175 200 0 20 40 60 80 100 120 140 160 0 10 20 30 40 50 60 70
Valid RMSE
Valid RMSE
Valid RMSE
0.8 0.8
0.8 Ours with MLP Ours with MLP Ours with MLP 0.8
Ours with MLP
Train Data 0.7 Train Data 0.7 Train Data 0.7
Train Data
0.7
Test Data Test Data Test Data Test Data
0.6 0.6
0.6
0.6
0.5 0.5
0.5
0.5
0.4 0.4
0.4
0.4
0.3 0.3
0.3
Valid Accuracy
Valid Accuracy
Valid Accuracy
0.55 0.7
0.85
0.90
0.6
0.80
our sacling law 0.85
our sacling law 0.50 our sacling law our sacling law
M1 M1 M1 0.5
M1
M2 M2 0.45
M2 M2
0.75 M3 0.80
M3 M3 0.4
M3
M4 M4 M4 M4
0.70
Ours with MLP 0.75
Ours with MLP 0.40
Ours with MLP 0.3
Ours with MLP
Train Data Train Data Train Data Train Data
Test Data 0.70
Test Data 0.35 Test Data 0.2
Test Data
Epoch Epoch Epoch Epoch
0 10 20 30 40 50 60 70 80 0 20 40 60 80 100 0 5 10 15 20 25 30 35 0 20 40 60 80
Valid RMSE
1.00
Valid RMSE
Valid RMSE
Ours with MLP Ours with MLP 0.95 Ours with MLP 0.8
Ours with MLP
0.8 Train Data Train Data Train Data 0.7
Train Data
Test Data 0.95
Test Data 0.90 Test Data Test Data
0.7 0.6
0.90 0.85
0.5
0.6
0.80
0.85
0.4
0.5
0.75
0.80 0.3
Figure 6: The visualization results of fitting the training dynamics (the validation curves of an MLP
trained with default hyper-parameters) on 16 datasets. The datasets in the first two and last two rows
are DNN-friendly and Tree-friendly, respectively. The first and the third rows represent classification
tasks, while the second and the fourth rows represent regression tasks.
We equip representative tabular methods with various encoding strategies and report their average
ranks on the tiny benchmark 1 in Table 3 in the main paper. In this section, we show more detailed
results in Table 9, Table 10, Table 11, and Table 12, which contain the results when we equip encoding
strategies with deep tabular methods and classical methods on classification and regression tasks,
respectively.
25
Table 8: The detailed results of Performance-Efficiency-Size comparison in Figure 1 in the main
paper. We list the average rank, running time, and checkpoint size of each method. “Dummy”
means the naive baseline. “LR”, “XGB”, “CatB”, and “RForest” denote “Logistic regression/Linear
Regression”, “XGBoost”, “CatBoost”, and “Random Forest”, respectively. TabPFN cannot be applied
to regression tasks.
Binary Multi-Class Regression All tasks
avg. size avg. rank avg. time avg. rank avg. time avg. rank avg. time avg. rank avg. time
Dummy 0 KB 11.96 1.26 × 10−4 12.44 1.08 × 10−4 11.27 7.25 × 10−5 11.81 1.00 × 10−4
LR 1.146 KB 8.06 1.21 × 10−1 8.66 2.26 × 10−1 8.60 1.22 × 10−2 8.43 1.06 × 10−1
KNN 2775 KB 8.94 2.96 × 10−1 7.95 3.24 × 10−1 7.14 4.91 × 10−1 7.96 3.81 × 10−1
SVM 1730 KB 8.57 1.09 × 100 9.91 2.45 × 100 9.64 3.23 × 100 9.35 2.30 × 100
XGB 12535 KB 4.79 4.39 × 100 5.33 1.67 × 101 4.13 9.79 × 10−1 4.67 2.31 × 100
CatB 18290 KB 4.24 1.47 × 101 4.44 3.11 × 101 3.15 9.05 × 100 3.86 1.68 × 101
RForest 182266 KB 6.12 1.22 × 100 6.56 1.15 × 100 5.25 8.61 × 100 5.89 4.13 × 100
MLP 2180 KB 6.63 1.58 × 101 5.99 1.51 × 101 6.26 1.42 × 101 6.31 1.50 × 101
ResNet 4383 KB 6.86 2.64 × 101 5.45 2.84 × 101 6.24 2.40 × 101 6.24 2.60 × 101
FT-T 7330 KB 5.53 5.89 × 101 5.49 4.85 × 101 5.03 9.34 × 101 5.32 6.98 × 101
DCNv2 10250 KB 6.95 2.55 × 101 6.21 2.61 × 101 6.29 2.02 × 101 6.49 2.36 × 101
TabR 14410 KB 4.73 1.81 × 102 3.71 2.53 × 102 4.95 1.69 × 102 4.55 1.95 × 102
TabPFN 100928 KB 6.39 7.33 × 10−4 6.98 6.59 × 10−4 - - - -
Table 9: Results of deep methods using different encoding strategies on classification tasks in Tiny
Benchmark 1. We report the average accuracies over 15 seeds, with each method tuned for 100
iterations. ID continues the ID in Table 4. We use “TF”, “DF”, and “Tie” to denote the subsets of
datasets that are tree-friendly, DNN-friendly, and those without any preference, respectively.
MLP ResNet FT-T TabR
ID Type Vanilla w/ E-Q w/ E-T Vanilla w/ E-Q w/ E-T Vanilla w/ E-Q w/ E-T Vanilla w/ E-Q w/ E-T
300 TF .6031 .5901 .5787 .6034 .5874 .5993 .5982 .5879 .5836 .5888 .5722 .5847
200 TF .8581 .8477 .8611 .8594 .8526 .8588 .8534 .8463 .8534 .8656 .8667 .8699
193 TF .9783 .9786 .9789 .9786 .9777 .9776 .9789 .9789 .9788 .9796 .9786 .9783
296 TF .7268 .7369 .7273 .7259 .7581 .7144 .7234 .7295 .7249 .7590 .7606 .7486
171 TF .8550 .8756 .8648 .8522 .8830 .8733 .8751 .8934 .8791 .9319 .9430 .9210
234 TF .9669 .9683 .9632 .9668 .9713 .9689 .9671 .9666 .9704 .9668 .9671 .9679
196 TF .8012 .9310 .9308 .7680 .9310 .9308 .7973 .9309 .9309 .9151 .9310 .9309
56 Tie .7822 .7716 .7644 .7684 .7807 .7473 .7702 .7798 .7411 .7873 .7776 .7638
88 Tie .8672 .9057 .9093 .9227 .9015 .9067 .9367 .9422 .9385 .9568 .9220 .9065
151 Tie .7248 .7251 .7305 .7473 .7357 .7587 .7664 .7724 .7691 .8697 .8688 .8720
194 Tie .9574 .9921 .9370 .9680 .9995 .9079 1.000 .9991 1.000 .9497 .9973 .9549
42 Tie .6692 .6733 .6730 .6696 .6736 .6733 .6728 .6720 .6708 .6712 .6708 .6743
226 Tie .7338 .7310 .7323 .7333 .7329 .7317 .7500 .7489 .7485 .7297 .7316 .7326
231 DF .9825 .9799 .9803 .9864 .9819 .9840 .9779 .9757 .9778 .9852 .9875 .9856
279 DF .6956 .6782 .6815 .6977 .6710 .6712 .6932 .6865 .6888 .6924 .6801 .6812
50 DF .6985 .6534 .5316 .7092 .6642 .6040 .6184 .6309 .5310 .7104 .7027 .7039
199 DF .9527 .9722 .9719 .8696 .9277 .9139 .9761 .9869 .9797 .9852 .9917 .9950
240 DF .9854 .9830 .9825 .9885 .9833 .9847 .9914 .9870 .9861 .9886 .9791 .9802
29 DF .8334 .8321 .8332 .8328 .8324 .8333 .8333 .8315 .8331 .8334 .8334 .8332
48 DF .7304 .6536 .5451 .7602 .6249 .5822 .6703 .6665 .5379 .7784 .7224 .6616
173 DF .9494 .9735 .9488 .9472 .9614 .9491 .9713 .9748 .9628 .9794 .9753 .9710
65 DF .8626 .8544 .8601 .8626 .8483 .8463 .8723 .8721 .8703 .8415 .8501 .8472
5 DF .3902 .3413 .3454 .3861 .3686 .8463 .3678 .4124 .4051 .3714 .3768 .3416
Based on the results in Table 9 and Table 10, we find that while the PLE encoding generally benefits
deep tabular methods, it is particularly effective on tree-friendly datasets. However, it can occasionally
have adverse effects on DNN-friendly datasets.
26
Table 10: Results of deep methods using different encoding strategies on regression tasks. We report
the average RMSE over 15 seeds, with each method tuned for 100 iterations. ID continues the ID
in Table 4. We use “TF”, “DF”, and “Tie” to denote the subsets of datasets that are tree-friendly,
DNN-friendly, and those without any preference, respectively. The scientific notation beside the
dataset name indicates the scale of the results, e.g., ×10 means the final value of all results should be
multiple by 10.
MLP ResNet FT-T TabR
ID Type Vanilla w/ E-Q w/ E-T Vanilla w/ E-Q w/ E-T Vanilla w/ E-Q w/ E-T Vanilla w/ E-Q w/ E-T
73 (×102 ) TF .1325 .1326 .1304 .1322 .1322 .1384 .1320 .1300 .0917 .1319 .1301 .1037
7 (×10−3 ) TF .1563 .1539 .1572 .1572 .1579 .1545 .1554 .1540 .1509 .1531 .1566 .1562
133 (×101 ) TF .2938 .2878 .2878 .2972 .2962 .2889 .2964 .2901 .2942 .3143 .3130 .3145
25 (×104 ) TF .1408 .1716 .1407 .1372 .1389 .1354 .1382 .1341 .1383 .1272 .1355 .1476
76 (×102 ) TF .7806 .8059 .7929 .7643 .7978 .7869 .7633 .7335 .7500 .7497 .7478 .7470
106 (×102 ) Tie .2874 .2876 .2860 .2926 .2927 .2927 .2868 .2867 .2859 .2887 .2867 .2869
114 (×100 ) Tie .1011 .1019 .1024 .1022 .1030 .1025 .1023 .1032 .1034 .1036 .1037 .1028
41 (×10−4 ) Tie .1228 .0764 .0998 .0826 .0840 .0788 .0533 .0534 .0523 .0522 .0526 .0522
70 (×100 ) Tie .2369 .2519 .2536 .2194 .2517 .2382 .2137 .2251 .2219 .2371 .2461 .2457
116 (×100 ) Tie .4595 .5033 .5088 .4586 .5014 .4967 .4563 .4898 .5062 .4515 .5346 .5503
21 (×100 ) Tie .6013 .7704 .6457 .5314 .5418 .5816 .5398 .5788 .5648 .4708 .4972 .4960
66 (×10−1 ) DF .2844 .7450 .8665 .2651 .3432 .5252 .2959 .2731 .2954 .2381 .2406 .2421
242 (×101 ) DF .2578 .2822 .2615 .3102 .2848 .2830 .2216 .2414 .2480 .2352 .3026 .3208
241 (×101 ) DF .2436 .2784 .2656 .3091 .3030 .2823 .2158 .2312 .2503 .2442 .2960 .3055
32 (×100 ) DF .2433 .3006 .3014 .2292 .2859 .2682 .2839 .2922 .3117 .2098 .2696 .2981
90 (×102 ) DF .1552 .1569 .1577 .1546 .1560 .1577 .1530 .1543 .1539 .1566 .1538 .1529
261 (×102 ) DF .2802 .2927 .2452 .2889 .2603 .2838 .1346 .1481 .1356 .2760 .2379 .2163
Table 11: Results of classical methods using different encoding strategies on classification tasks in
Tiny Benchmark 1. We report the average accuracies over 15 seeds, with each method tuned for 100
iterations. ID continues the ID in Table 4. We use “TF”, “DF”, and “Tie” to denote the subsets of
datasets that are tree-friendly, DNN-friendly, and those without any preference, respectively.
XGBoost Random Forest SVM Catboost
ID Type Vanilla w/ E-Q w/ E-T Vanilla w/ E-Q w/ E-T Vanilla w/ E-Q w/ E-T Vanilla w/ E-Q w/ E-T
300 TF .5984 .6081 .6123 .6247 .6056 .6018 .5859 .5791 .6027 .6083 .5971 .6173
200 TF .8627 .8673 .8660 .8698 .8654 .8641 .8507 .8555 .8649 .8624 .8621 .8624
193 TF .9789 .9782 .9782 .9803 .9791 .9788 .9798 .9804 .9789 .9794 .9795 .9786
296 TF .7500 .7281 .7425 .7386 .7342 .7352 .7045 .7260 .7136 .7534 .7470 .7280
171 TF .9246 .9306 .9315 .8577 .8436 .8370 .7527 .8091 .8049 .9169 .9313 .9250
234 TF .9745 .9746 .9762 .9703 .9719 .9738 .9565 .9616 .9689 .9667 .9669 .9712
196 TF .9298 .9310 .9310 .9311 .9311 .9310 .7954 .9259 .9306 .9308 .9310 .9310
56 Tie .7500 .7489 .7420 .7738 .7742 .7622 .7800 .7867 .7989 .7811 .7731 .7869
88 Tie .9090 .9093 .9160 .8760 .9220 .8847 .8348 .7977 .8067 .9357 .9345 .9160
151 Tie .7675 .7404 .7495 .7380 .7092 .7178 .6812 .6979 .7008 .7637 .7619 .7693
194 Tie 1.000 1.000 1.000 1.000 .9923 1.000 1.000 1.000 .9422 1.000 1.000 1.000
42 Tie .6723 .6763 .6791 .6650 .6650 .6708 .6286 .6460 .6505 .6672 .6736 .6768
226 Tie .7514 .7473 .7460 .7407 .7367 .7377 .7237 .7318 .7286 .7508 .7510 .7511
231 DF .9824 .9765 .9829 .9801 .9837 .9816 .9582 .9662 .9689 .9824 .9815 .9821
279 DF .6921 .6996 .6880 .6929 .6796 .6789 .6971 .6839 .6839 .6924 .6898 .6900
50 DF .6023 .6241 .6121 .5653 .5541 .5447 .7041 .6998 .6256 .6158 .6305 .6164
199 DF .8610 .8590 .8600 .8262 .8263 .8286 .6801 .7483 .7507 .8647 .8634 .8637
240 DF .9824 .9846 .9848 .9732 .9729 .9805 .8815 .9535 .9539 .9860 .9861 .9858
29 DF .8325 .8325 .8321 .8325 .8328 .8330 .7921 .7180 .8332 .8326 .8327 .8326
48 DF .6487 .6489 .6271 .6416 .6245 .6056 .7664 .7252 .6452 .6558 .6365 .6638
173 DF .9474 .9628 .9633 .9346 .9685 .9526 .7634 .8747 .8629 .9469 .9556 .9576
65 DF .8478 .8485 .8431 .8537 .8476 .8499 .8683 .8435 .8537 .8488 .8558 .8567
5 DF .4301 .4350 .4404 .4108 .4247 .4090 .3465 .3707 .3629 .3815 .4106 .3959
27
Table 12: Results of classical methods using different encoding strategies on regression tasks. We
report the average RMSE over 15 seeds, with each method tuned for 100 iterations. ID continues the
ID in Table 4. We use “TF”, “DF”, and “Tie” to denote the subsets of datasets that are tree-friendly,
DNN-friendly, and those without any preference, respectively. The scientific notation beside the
dataset name indicates the scale of the results, e.g., ×10 means the final value of all results should be
multiple by 10.
XGBoost Random Forest SVM Catboost
ID Type Vanilla w/ E-Q w/ E-T Vanilla w/ E-Q w/ E-T Vanilla w/ E-Q w/ E-T Vanilla w/ E-Q w/ E-T
73 (×102 ) TF .1053 .0743 .0717 .0766 .0764 .0759 .1326 .4070 .3891 .1024 .0806 .8090
7 (×10−3 ) TF .1527 .1525 .1519 .1559 .1560 .1560 .1696 .1767 .1664 .1465 .1473 .1462
133 (×101 ) TF .2980 .2800 .2813 .2736 .2738 .2740 .2955 .3003 .3007 .2924 .2922 .2902
25 (×104 ) TF .1199 .1180 .1171 .1228 .1228 .1228 .1552 .1573 .1568 .1275 .1239 .1253
76 (×103 ) TF .0718 .0708 .0712 .0760 .0761 .0762 .1561 .1565 .1559 .0726 .0730 .7318
106 (×102 ) Tie .2863 .2849 .2847 .2936 .2938 .2936 .2859 .2859 .2853 .2864 .2850 .2853
114 (×100 ) Tie .1010 .1009 .1023 .1023 .1023 .1022 .1059 .1026 .1013 .1044 .1017 .1018
41 (×10−2 ) Tie .0533 .0529 .0532 .0555 .0556 .0556 .1507 .1423 .1340 .0525 .0530 .0531
70 (×100 ) Tie .2218 .2196 .2197 .2293 .2294 .2298 .2796 .2678 .2893 .2135 .2152 .2177
116 (×100 ) Tie .4589 .4768 .4644 .4565 .4565 .4563 .5822 .5027 .5519 .4705 .4712 .4695
21 (×101 ) Tie .0648 .0642 .0630 .0798 .0798 .0798 .1038 .1111 .1087 .0537 .0539 .0549
66 (×10−1 ) DF .4063 .4536 .4003 .4478 .4486 .4429 .5894 .4279 .4282 .3483 .3401 .3350
242 (×102 ) DF .0438 .0427 .0450 .0498 .0497 .0497 .3095 .2378 .2337 .3669 .3603 .3650
241 (×102 ) DF .0478 .0468 .0457 .0542 .0537 .0538 .3057 .2380 .2315 .0419 .0411 .0413
32 (×100 ) DF .2738 .2772 .2748 .2945 .2939 .2932 .3564 .3206 .3981 .2674 .2711 .2735
90 (×102 ) DF .1559 .1566 .1563 .1560 .1564 .1560 .1750 .1571 .1571 .1582 .1561 .1592
261 (×102 ) DF .2389 .2184 .2054 .2501 .2499 .2534 .1935 .2858 .2403 .2463 .2353 .2216
We release the mean and standard deviation of the performance metric results (accuracy for classifica-
tion tasks and RMSE for regression tasks) for each dataset-method pair, organized into six tables by
task type.
For classification tasks, we test the performance of following 27 models: Dummy, Logistic Regression,
NCM, Naive Bayes, KNN, SVM, XGBoost, CatBoost, Random Forest, LightGBM, TabPFN, MLP,
ResNet, NODE, SwitchTab, TabNet, TabCaps, TANGOS, DANets, FT-T, AutoInt, DCNv2, SNN,
TabTransformer, PTaRL, GrowNet, and TabR. The mean and the standard deviation of accuracy on
101 binary classification tasks are shown in Table 13 and Table 16, indexed by their IDs in Table 4.
Similarly, The results of 80 multi-class classification tasks are shown in Table 14 and Table 17.
For regression tasks, we test the performance of following 24 models: Dummy, Linear Regression,
KNN, SVM, XGBoost, CatBoost, Random Forest, LightGBM, MLP, ResNet, NODE, SwitchTab,
TabNet, TANGOS, DANets, FT-T, AutoInt, DCNv2, SNN, TabTransformer, PTaRL, GrowNet, TabR,
and DNNR. The mean and the standard deviation of RMSE on 119 regression tasks are shown in
Table 15 and Table 18, indexed by their IDs in Table 4. RMSE has a wide range of distributions
across different datasets, for presentation purposes, we choose the scaling based on the median, and
denote excessively large values as “INF”.
28
Table 13: Additional results of binary classification datasets. We list the average accuracy of each
dataset-method pair on 15 random seeds. ID continues the ID in Table 4. “DU” means Dummy as the
naive baseline, “LR”, “NB”, “XGB”, “CB”, “RF”, “LG”, “PFN”, “RN”, “ND”, “ST”, “TN”, “TC”,
“TG”, “DAN”, “FT”, “AI”, “DCN”, “TT”, “PT”, “GN”, and “TR” denote “Logistic Regression”,
“Naive Bayes”, “XGBoost”, “CatBoost”, “Random Forest”, “LightGBM”, “TabPFN”, “ResNet”,
“NODE”, “SwitchTab”, “TabNet”, “TabCaps”, “TANGOS”, “DANets”, “FT-T”, “AutoInt”, “DCNv2”,
“TabTransformer”, “PTaRL”, “GrowNet”, and “TabR”, respectively. In each cell, if the experiment
fails to perform due to being out of memory, we denote it by “OOM”. TabPFN cannot handle datasets
with more than 10 classes or more than 100 features.
ID DU LR NCM NB KNN SVM XGB CB RF LG PFN MLP RN ND ST TN TC TG DAN FT AI DCN SNN TT PT GN TR
8 .9422 .9422 .5021 .9422 .9411 .9422 .9428 .9467 .9426 .9447 .9422 .9432 .9436 .9422 .9421 .9418 .9422 .9421 .9425 .9420 .9431 .9421 .9431 .9450 .9431 .9378 .9426
11 .6560 .9782 .9669 .9552 .9837 .9781 .9876 .9878 .9852 .9873 .9827 .9846 .9846 .9856 .9818 .9834 .9834 .9838 .9840 .9860 .9848 .9841 .9848 .9594 .9840 .9839 .9868
17 .6529 .6527 .6021 .6855 .7779 .6407 .8099 .8133 .7979 .8075 .7599 .8003 .8041 .8117 .7703 .7974 .7981 .7992 .8009 .8131 .8100 .8153 .8150 .8139 .8112 .8112 .8045
18 .7965 .8160 .6975 .8395 .8485 .8073 .8688 .8759 .8681 .8676 .8579 .8626 .8654 .8654 .8495 .8631 .8650 .8659 .8636 .8709 .8674 .8674 .8682 .8313 .8646 .8624 .8762
19 .6194 .7052 .6306 .6082 .6679 .7082 .6858 .6761 .6891 .6853 .7042 .6975 .7035 .6542 .6903 .6577 .7052 .7005 .7075 .7015 .7002 .7012 .6958 .6928 .7085 .6694 .7050
23 .8607 .8646 .7486 .7745 .8641 .8642 .8676 .8678 .8665 .8678 .8643 .8680 .8680 .8678 .8674 .8669 .8679 .8677 .8680 .8676 .8675 .8681 .8676 .8652 .8681 .8651 .8674
27 .5002 .8406 .7485 .7362 .8522 .8413 .9000 .9082 .8765 .9070 .8731 .8847 .8851 .8730 .8135 .8787 .8812 .8835 .8830 .8923 .8847 .8858 .8843 .8501 .8805 .8743 .9135
28 .5003 .7108 .6403 .5869 .6775 .6893 .7331 .7340 .7325 .7335 .7254 .7339 .7314 .7320 .7284 .7147 .7316 .7298 .7323 .7322 .7326 .7321 .7328 .6799 .7321 .6677 .7314
29 .8315 .8323 .5228 .8332 .8325 .7921 .8325 .8326 .8325 .8332 .8315 .8334 .8328 .8331 .8325 .8319 .8333 .8323 .8321 .8333 .8330 .8327 .8332 .8320 .8331 .8326 .8322
37 .8504 .8884 .7567 .8415 .8705 .8815 .8754 .8905 .8801 .8783 .8932 .8787 .8756 .8580 .8662 .8530 .8854 .8743 .8679 .8927 .8908 .8917 .8826 .8824 .8876 .8692 .8609
40 .5325 .7273 .5671 .6277 .6494 .7227 .7261 .7414 .7351 .7328 .7261 .7391 .7636 .5772 .6349 .6087 .7065 .7391 .7388 .7319 .7160 .7671 .7330 .7177 .7322 .6644 .7576
42 .5968 .6255 .6414 .6455 .6455 .6286 .6723 .6672 .6650 .6779 .6502 .6692 .6696 .6634 .6597 .6628 .6666 .6640 .6619 .6728 .6650 .6627 .6584 .6292 .6655 .6476 .6710
43 .6563 .6971 .6660 .6799 .8195 .7079 .8462 .8500 .8517 .8362 .8501 .8458 .8482 .8438 .7922 .8157 .8412 .8510 .8489 .8477 .8486 .8497 .8465 .7703 .8379 .8460 .8384
44 .5205 .7423 .7225 .7063 .7251 .7443 .7411 .7427 .7390 .7424 .7382 .7391 .7388 .7451 .7319 .7310 .7411 .7356 .7361 .7410 .7393 .7374 .7342 .7353 .7340 .7275 .7396
45 .5068 .7520 .5095 .4850 .6213 .7391 .6714 .6816 .6750 .6812 .7500 .7384 .7480 .5577 .6625 .6787 .7322 .7315 .7277 .7346 .7090 .7370 .7193 .7386 .7341 .7010 .7428
46 .5147 .7062 .5043 .5055 .5765 .7034 .6381 .6630 .5849 .6642 .6989 .7121 .7121 .5900 .6539 .7106 .7053 .6725 .7112 .6901 .7089 .7108 .7114 .7070 .7123 .7010 .7116
47 .5068 .7520 .5068 .5286 .5940 .7520 .6627 .6398 .6554 .6618 .7559 .6901 .7464 .6523 .5232 .5135 .6111 .5010 .5370 .5179 .6567 .7480 .7262 .7424 .6191 .6754 .7520
48 .5150 .7738 .5395 .5204 .5913 .7664 .6487 .6558 .6416 .6460 .7589 .7304 .7602 .5506 .6118 .5671 .6957 .7052 .6716 .6703 .6883 .7450 .7402 .7313 .7199 .6941 .7655
49 .5150 .7221 .5293 .5292 .5480 .7142 .6407 .6495 .5818 .6440 .6725 .7173 .7178 .5436 .5387 .5869 .6875 .7129 .7058 .6913 .7006 .7132 .7138 .7132 .7088 .6547 .7202
50 .5148 .7061 .5216 .5217 .5254 .7041 .6023 .6158 .5653 .6228 .5813 .6985 .7092 .5225 .5146 .6413 .6540 .5270 .6562 .6184 .6884 .6909 .6899 .6699 .6512 .6336 .7080
51 .5296 .7056 .5046 .5052 .5313 .6913 .6080 .6165 .5657 .6225 .5451 .6831 .6932 .5347 .5329 .6762 .6216 .5362 .6494 .6717 .6559 .6976 .6882 .6965 .6923 .6214 .7109
52 .5014 .7057 .5259 .4877 .5204 .7139 .6025 .5889 .5713 .5913 .7308 .7204 .7005 .7063 .5317 .5671 .6946 .6966 .7014 .6665 .6767 .7152 .5926 .7150 .5955 .6545 .7124
53 .5179 .7100 .5200 .5140 .5461 .7099 .6121 .6259 .5718 .6272 .6290 .7061 .7096 .5338 .5334 .6243 .6838 .5885 .6871 .6880 .6971 .7045 .6990 .7034 .7046 .6586 .6432
56 .6967 .8000 .7367 .7933 .7900 .7800 .7500 .7811 .7738 .7733 .7993 .7822 .7684 .7862 .7720 .7511 .7924 .7882 .7864 .7702 .7624 .7693 .7704 .7458 .7816 .7538 .7871
58 .5000 .5125 .5062 .4938 .5000 .5656 .6096 .6808 .5579 .6096 .5894 .5846 .5777 .5696 .5377 .4938 .6012 .5710 .5633 .4931 .5413 .5890 .6365 .6433 .5131 .5775 .5894
59 .5000 .4594 .4594 .4594 .5469 .4979 .6665 .6796 .6502 .6823 .6556 .6400 .6481 .5329 .5427 .4954 .6704 .6279 .6373 .5325 .5942 .6637 .6679 .6592 .5981 .6483 .6685
63 .7508 .7688 .7020 .7581 .7811 .7616 .7974 .7951 .7939 .7967 .7735 .7835 .7817 .7910 .7753 .7832 .7871 .7837 .7855 .7980 .7960 .7972 .7922 .7968 .7977 .7732 .7906
65 .8401 .8673 .6769 .7415 .8401 .8683 .8478 .8488 .8537 .8467 .8707 .8626 .8626 .8408 .8540 .8363 .8565 .8628 .8692 .8723 .8689 .8746 .8730 .8549 .8449 .8630 .8766
67 .6724 .8074 .7515 .4629 .8707 .8062 .8972 .8921 .8821 .8962 .8337 .8739 .8716 .8443 .8365 .8620 .8728 .8735 .8729 .8850 .8719 .8746 .8718 .8530 .8779 .7841 .8933
75 .5000 .7388 .6511 .6267 .7193 .7388 .8163 .8103 .8134 .8101 .7682 .7663 .7632 .7741 .7516 .7684 .7676 .7552 .7596 .8038 .7651 .7565 .7640 .7475 .7606 .7168 .7854
83 .8455 .8848 .8515 .1788 .8576 .8879 .9105 .8871 .8836 .9026 - .8828 .8725 .8491 .8693 .8406 .8705 .8800 .8788 .8968 .8778 .8909 .8885 .8909 .8848 .8697 .8929
84 .6483 .7976 .7821 .7358 .8465 .7934 .8763 .8787 .8690 .8788 .8669 .8724 .8761 .8701 .8512 .8698 .8725 .8691 .8691 .8759 .8703 .8718 .8688 .8309 .8698 .8670 .8813
85 .8504 .8839 .7277 .8192 .8728 .8708 .8827 .8805 .8763 .8802 .8888 .8704 .8783 .8615 .8754 .8494 .8862 .8760 .8845 .8688 .8826 .8643 .8885 .8772 .8808 .8695 .8772
91 .8399 .8355 .7368 .8311 .8289 .8319 .8423 .8427 .8507 .8364 .8338 .8303 .8256 .8425 .8266 .8377 .8358 .8298 .8330 .8363 .8383 .8300 .8304 .8308 .8386 .8320 .8344
95 .5568 .9263 .9132 .5984 .9584 .9290 .9731 .9679 .9624 .9734 .9495 .9703 .9716 .9486 .9445 .9495 .9681 .9708 .9699 .9676 .9694 .9696 .9709 .9730 .9698 .9620 .9760
97 .8750 .8750 .7824 .6852 .8750 .8889 .8753 .8750 .8731 .8694 .8704 .8765 .8660 .8747 .8670 .8133 .8744 .8756 .8759 .8710 .8747 .8586 .8654 .8731 .8747 .8614 .8744
98 .6494 .7597 .7662 .7468 .7143 .7468 .7645 .7775 .7727 .7814 .7558 .7606 .7494 .7706 .7398 .7303 .7524 .7710 .7251 .7558 .7662 .7364 .7649 .6861 .7580 .7602 .7597
99 .8804 .8947 .7560 .2201 .8756 .9027 .8561 .8775 .8743 .8807 .8794 .8855 .8788 .8801 .8797 .8705 .8804 .8813 .8817 .8826 .8801 .8762 .8759 .8928 .8797 .8769 .8772
100 .5200 .8720 .8480 .8580 .8640 .8587 .8716 .8725 .8740 .8747 .8745 .8691 .8677 .8607 .8513 .8613 .8663 .8661 .8508 .8672 .8707 .8717 .8649 .8639 .8651 .8633 .8687
101 .6635 .8531 .7915 .7014 .8910 .8442 .8506 .8730 .8742 .8408 .8752 .8667 .8667 .6976 .8610 .7987 .8793 .8714 .8686 .8695 .8512 .8708 .8771 .8626 .8537 .8569 .8603
104 .9853 .9922 .9422 .9324 .9912 .9912 .9921 .9920 .9913 .9902 .9907 .9913 .9917 .9890 .9913 .9893 .9915 .9920 .9922 .9920 .9909 .9911 .9912 .9910 .9913 .9914 .9918
107 .5002 .7253 .6984 .6920 .7236 .7234 .7387 .7399 .7308 .7383 .7256 .7390 .7393 .7401 .7330 .7384 .7391 .7382 .7390 .7403 .7396 .7391 .7391 .7370 .7387 .7337 OOM
111 .8546 .8681 .6867 .8801 .8951 .8756 .9557 .9517 .9544 .9539 .9330 .9162 .9281 .9129 .8865 .8932 .9172 .9225 .9231 .9519 .9258 .9196 .9129 .8574 .9173 .9085 .9418
115 .6098 .6098 .5152 .6174 .6387 .6098 .6144 .6250 .6293 .6286 .6154 .6301 .6280 .6505 .6143 .6066 .6309 .6271 .6258 .6310 .6213 .6258 .6115 .6235 .6313 .6132 .6263
117 .9461 .9658 .6580 .9057 .9658 .9665 .9883 .9879 .9836 .9869 .9914 .9885 .9842 .9461 .9484 .9711 .9825 .9858 .9846 .9849 .9894 .9851 .9835 .9761 .9789 .9804 .9872
122 .7518 .8337 .7663 .6759 .8289 .8361 .8553 .8566 .8476 .8512 .8442 .8434 .8445 .8275 .8382 .8302 .8415 .8468 .8407 .8480 .8339 .8405 .8308 .8446 .8394 .8286 .8483
123 .7514 .8280 .7141 .5082 .7974 .8307 .8362 .8451 .8454 .8382 .8405 .8348 .8292 .8134 .8278 .8025 .8356 .8393 .8263 .8369 .8364 .8325 .8339 .8180 .8372 .8238 .8357
124 .7514 .8335 .7349 .8072 .8083 .8150 .8501 .8500 .8483 .8442 .8400 .8356 .8373 .8155 .8304 .8267 .8413 .8313 .8376 .8483 .8445 .8476 .8364 .8148 .8483 .8138 .8420
126 .5550 .5975 .5750 .5775 .5850 .6042 .6140 .6288 .6120 .6047 .5840 .5867 .5597 .5865 .5665 .5522 .5842 .5755 .5820 .6032 .6005 .6097 .6143 .6363 .6103 .5988 .5622
139 .8830 .8905 .8002 .8336 .8945 .8887 .9070 .9082 .9041 .9071 .8926 .9028 .9034 .9035 .8988 .8999 .9016 .8998 .9021 .9095 .9069 .9050 .9070 .8999 .9060 .9025 .9085
142 .5564 .5309 .5418 .5673 .5564 .5309 .4812 .5185 .4732 .5641 .5564 .5333 .5287 .5331 .5358 .5171 .5450 .5505 .5513 .5476 .5522 .5435 .5212 .5455 .5513 .5275 .5404
146 .8590 .8680 .7360 .8790 .8970 .8820 .9557 .9575 .9552 .9550 .9353 .9345 .9387 .9332 .8983 .8997 .9418 .9263 .9312 .9634 .9551 .9479 .9479 .8929 .9422 .9215 .9565
150 .9677 .9663 .8761 .2265 .9677 .9679 .9675 .9689 .9700 .9713 .9700 .9687 .9673 .9709 .9678 .9674 .9681 .9680 .9699 .9665 .9687 .9684 .9675 .9687 .9677 .9671 .9647
151 .5002 .6783 .6651 .6530 .7447 .6812 .7675 .7637 .7380 .7422 .6936 .7248 .7473 .6981 .6869 .6907 .7008 .7090 .7239 .7664 .7592 .7359 .7264 .7501 .7508 .6559 .8800
157 .5001 .7407 .6019 .5537 .6787 .7098 .7817 .7796 .7795 .7782 .7693 .7580 .7540 .6225 .7319 .7469 .7567 .7546 .7554 .7588 .7561 .7556 .7528 .7467 .7550 .7443 .7755
159 .5473 .8285 .7941 .8128 .9343 .8288 .9596 .9631 .9421 .9603 .9132 .9606 .9620 .9557 .9390 .9543 .9578 .9586 .9584 .9628 .9610 .9582 .9582 .9553 .9577 .9505 .9611
160 .5391 .6126 .5968 .5952 .6128 .6091 .6436 .6462 .6334 .6426 .6145 .6277 .6316 .6391 .6152 .6253 .6313 .6215 .6242 .6387 .6406 .6410 .6422 .6247 .6419 .6376 .6265
163 .7788 .8083 .6122 .5662 .8068 .8140 .8260 .8244 .8236 .8270 .8210 .8241 .8248 .8249 .8157 .8215 .8236 .8238 .8259 .8255 .8249 .8253 .8235 .8246 .8249 .8213 .8237
164 .5309 .9425 .9173 .9362 .9467 .9431 .9502 .9482 .9489 .9485 .9472 .9449 .9473 .9444 .9408 .9435 .9448 .9460 .9469 .9471 .9473 .9467 .9442 .9068 .9461 .9460 .9482
166 .9841 .9841 .5921 .2583 .9841 .9841 .9864 .9866 .9841 .9906 .9841 .9848 .9851 .9841 .9832 .9837 .9830 .9849 .9858 .9876 .9848 .9839 .9850 .9841 .9841 .9840 .9885
170 .5511 .6222 .5607 .4606 .9222 .6248 .9449 .9568 .8841 .9460 .9284 .9569 .9361 .8195 .6516 .6312 .9500 .9522 .9456 .9535 .9526 .9564 .9573 .9170 .9208 .8156 .9819
171 .5754 .7553 .6610 .7230 .8433 .7527 .9246 .9169 .8577 .9238 .7945 .8550 .8522 .8318 .7804 .8192 .8325 .8465 .8518 .8751 .8605 .8555 .8515 .7949 .8554 .8284 .9652
175 .5000 .5650 .5664 .5414 .5802 .5558 .6325 .6152 .6057 .6393 .5784 .5708 .5754 .5611 .5582 .5645 .5706 .5760 .5745 .5945 .5831 .5741 .5764 .5673 .5745 .5579 .6595
182 .5086 .8674 .7781 .7997 .7853 .8646 .9235 .9381 .9038 .9361 - .9012 .9149 .9131 .8919 .8084 .9154 .9049 .8980 .8606 OOM .9147 .8948 .8915 .8924 .9034 .9295
183 .7900 .9041 .8356 .8950 .8447 .9041 .9132 .9105 .9221 .9279 .9081 .8938 .9059 .9184 .9068 .8904 .9178 .9011 .9205 .9288 .9440 .8992 .9452 .8435 .9169 .8831 .9172
185 .5000 .7130 .7135 .6980 .7145 .7125 .7233 .7230 .7198 .7221 .7175 .7181 .7175 .7205 .7188 .7121 .7171 .7175 .7194 .7217 .7195 .7087 .7163 .7187 .7174 .7140 .7217
186 .4979 .9547 .4774 .4856 .5556 .7868 .5778 .5166 .5498 .6027 .5163 .6549 .7764 .4919 .5191 .4982 .5874 .6532 .6126 .5081 .5487 .5953 .6269 .8944 .5339 .5841 .7004
187 .5000 .8239 .7324 .7480 .8632 .8146 .8826 .8847 .8788 .8808 .8711 .8717 .8697 .8723 .8453 .8503 .8640 .8726 .8686 .8747 .8721 .8685 .8662 .8626 .8713 .8632 .8702
193 .9084 .9802 .9712 .9399 .9799 .9798 .9789 .9794 .9803 .9781 .9780 .9783 .9786 .9754 .9780 .9783 .9789 .9790 .9780 .9789 .9783 .9786 .9782 .9785 .9790 .9791 .9785
194 .8469 .9966 .9286 .9966 .8810 1.000 1.000 1.000 1.000 1.000 .9986 .9574 .9680 .8449 .9381 .9279 .9760 .9694 .9565 1.000 .9365 .9780 .9723 .9227 .9460 .8612 1.000
195 .5684 .6279 .6216 .6141 .6882 .6208 .7611 .7622 .7441 .7626 .6439 .7082 .7155 .7225 .6338 .6777 .7135 .7163 .7096 .7570 .7431 .7490 .7570 .7602 .7453 .7399 .7406
198 .8066 .8098 .7409 .7997 .8167 .8129 .8144 .8084 .8111 .8089 .8132 .8125 .8095 .8131 .7984 .8094 .8136 .8130 .8078 .8141 .8135 .8109 .8100 .8131 .8110 .8124 .8165
200 .8460 .8555 .8246 .8294 .8602 .8507 .8627 .8624 .8698 .8744 .8662 .8581 .8594 .8502 .8575 .8411 .8648 .8575 .8540 .8534 .8626 .8553 .8488 .8607 .8494 .8596 .8474
201 .5000 .8642 .7466 .7293 .8372 .8581 .8836 .8774 .8827 .8870 .8599 .8496 .8516 .8328 .8420 .8469 .8464 .8475 .8489 .8487 .8458 .8507 .8507 .8496 .8474 .8423 .8760
209 .5032 .5924 .6083 .6019 .6783 .6035 .8632 .8643 .7412 .8638 - .6152 .5938 .7645 .6066 .5530 .6226 .5818 .5856 .6691 .5731 .5917 .6141 .5952 .5729 .5091 .7240
29
210 .9768 .9826 .9410 .9598 .9866 .9819 .9867 .9883 .9852 .9876 .9865 .9867 .9871 .9821 .9707 .9831 .9850 .9861 .9878 .9871 .9870 .9865 .9870 .9817 .9865 .9843 .9872
221 .6070 .8337 .8327 .8569 .9668 .8369 .9805 .9803 .9701 .9795 .9340 .9832 .9804 .9613 .9034 .9719 .9767 .9786 .9807 .9817 .9821 .9826 .9825 .8856 .9798 .8355 .9893
222 .6713 .8504 .7112 .6809 .9112 .7785 .9459 .9466 .9456 .9458 .9119 .9194 .9135 .9273 .8502 .9261 .9252 .8991 .9148 .9276 .9090 .9250 .8288 .8509 .9185 .9153 .9426
224 .5011 .9589 .9103 .5933 .9656 .9583 .9695 .9695 .9638 .9694 .9631 .9697 .9700 .9619 .9635 .9655 .9699 .9697 .9694 .9691 .9692 .9681 .9654 .9700 .9682 .9696 .9699
225 .6222 .9969 .7749 .9002 .8788 .9964 1.000 1.000 1.000 1.000 .9950 .9809 .9808 .9979 .9504 .9890 .9893 .9730 .9829 .9965 .9937 .9845 .9830 .9193 .9822 .9072 .9969
230 .8451 .8836 .8155 .7299 .8739 .8824 .8995 .9063 .9038 .9012 .8969 .8939 .8957 .8875 .8905 .8908 .8916 .8958 .8958 .9019 .8976 .9032 .8969 .8868 .8989 .8829 .8964
232 .9369 .9428 .6607 .6588 .9507 .9408 .9479 .9440 .9500 .9481 .9513 .9374 .9374 .9369 .9439 .9346 .9466 .9474 .9303 .9483 .9453 .9431 .9400 .9352 .9446 .9446 .9442
235 .9324 .9324 .8514 .8829 .9369 .9324 .9381 .9351 .9357 .9375 .9369 .9273 .9210 .9324 .9258 .9219 .9291 .9363 .9336 .9309 .9324 .9183 .9273 .9324 .9318 .9243 .9336
236 .8978 .8850 .7955 .8019 .8914 .8871 .8856 .8880 .8941 .8773 .8950 .8820 .8867 .8978 .8835 .8884 .8999 .8920 .8907 .8978 .8952 .8867 .8929 .8976 .8909 .8963 .8846
237 .8767 .9041 .7911 .8596 .8904 .9027 .9151 .9016 .8913 .9110 .9078 .8989 .8852 .8779 .8854 .8747 .9021 .8989 .9005 .8959 .9034 .8975 .9016 .8968 .8995 .8975 .8989
239 .7068 .7382 .7327 .7530 .8686 .7294 .8766 .8797 .8852 .8807 .8624 .8684 .8714 .8538 .7987 .8466 .8480 .8583 .8610 .8809 .8622 .8710 .8593 .8070 .8632 .8643 .8892
240 .5002 .8790 .7685 .7189 .9534 .8815 .9824 .9860 .9732 .9818 .9748 .9854 .9885 .9808 .9569 .9770 .9881 .9845 .9853 .9914 .9905 .9882 .9878 .9758 .9841 .7142 .9887
246 .6635 .8768 .7773 .6588 .8673 .8711 .8645 .8613 .8515 .8616 .9087 .8932 .8667 .6768 .8657 .7912 .8831 .8970 .8926 .8537 .8641 .8720 .8720 .8761 .8720 .8569 .8632
249 .5722 .9252 .9081 .9081 .9239 .9213 .9189 .9223 .9218 .9234 .9298 .9179 .9210 .9191 .9178 .9166 .9198 .9203 .9163 .9204 .9205 .9192 .9223 .9244 .7902 .9136 .9216
250 .5047 .7730 .7743 .9770 .7757 .7705 .9715 .9711 .9367 .9741 .9804 .9741 .9766 .9688 .9221 .9485 .9735 .9726 .8180 .9765 .9695 .9696 .9730 .5079 .9746 .9764 .9795
251 .5000 .5805 .5624 .5483 .6640 .5718 .7922 .7858 .7490 .7983 .6738 .6600 .6496 .6475 .6227 .6281 .6601 .6596 .6581 .7196 .6822 .6488 .6490 .6282 .6429 .6258 .8824
255 .9342 .9323 .7505 .8337 .9246 .9295 .9313 .9293 .9340 .9295 .9342 .9277 .9219 .9340 .9280 .9287 .9333 .9251 .9215 .9280 .9273 .9250 .9282 .9264 .9297 .9292 .9262
263 .6059 .9229 .8914 .8284 .9175 .9191 .9523 .9467 .9330 .9461 .9473 .9386 .9391 .9399 .9301 .9134 .9373 .9397 .9374 .9393 .9356 .9383 .9296 .9381 .9389 .9323 .9388
265 .6350 .8400 .8150 .7650 .8250 .8360 .8407 .8340 .8247 .8440 .8400 .8303 .8263 .7017 .8217 .7710 .8353 .8317 .8347 .8330 .8257 .8223 .8217 .8293 .8350 .8327 .8360
266 .7000 .7200 .7000 .7050 .7250 .7350 .7467 .7430 .7450 .7433 .7310 .7250 .7110 .7227 .7230 .6873 .7087 .7307 .7273 .7123 .7250 .7123 .7157 .7077 .7113 .6860 .7297
273 .4995 .9102 .8302 .8966 .8820 .9161 .9454 .9524 .9316 .9424 .9320 .9309 .9329 .9154 .9112 .9231 .9337 .9287 .9329 .9471 .9270 .9357 .8994 .9350 .9315 .9161 .9638
274 .9677 .9677 .8864 .2478 .9714 .9677 .9702 .9718 .9717 .9712 .9692 .9685 .9667 .9678 .9522 .9681 .9672 .9689 .9686 .9674 .9674 .9662 .9664 .9686 .9670 .9660 .9678
275 .7346 .7921 .7189 .7473 .7793 .7944 .7962 .8057 .7968 .8031 .7970 .7925 .7899 .7988 .7840 .7854 .7906 .7914 .7913 .8005 .7982 .8002 .7957 .7820 .7977 .7902 .7973
283 .5007 .9811 .9797 .9804 .9797 .9791 .9763 .9774 .9745 .9757 .9791 .9773 .9745 .9791 .9728 .9629 .9796 .9748 .9783 .9782 .9762 .9676 .9781 .5040 .9762 .6404 .9762
289 .8862 .8944 .6887 .7544 .9056 .8950 .9025 .8998 .9001 .9041 .8956 .9078 .9105 .9006 .9032 .8921 .8998 .9049 .9072 .9013 .9040 .9056 .9025 .9052 .9040 .9033 .9050
296 .5010 .7045 .6947 .7143 .7710 .7045 .7500 .7534 .7386 .7339 .7341 .7268 .7259 .7517 .6881 .7204 .7198 .7280 .7190 .7234 .7162 .7211 .7205 .6828 .7212 .7448 .7393
Table 14: Additional results of multi-class classification datasets. We list the average accuracy
of each dataset-method pair on 15 random seeds. The column names continue the denotation in
Table 13.
5 .2008 .3475 .1969 .2625 .3707 .3465 .4301 .3815 .4108 .4368 - .3902 .3861 .2780 .3514 .2566 .3701 .3480 .3699 .3678 .3462 .3588 .3423 .3184 .3596 .3591 .3874
10 .3430 .7281 .7156 .7156 .7596 .7366 .7455 .7445 .7422 .7399 .7322 .7344 .7355 .7297 .7244 .7341 .7329 .7317 .7311 .7339 .7332 .7343 .7356 .6316 .7321 .7351 .7569
12 .4262 .5227 .4668 .4825 .5629 .5185 .5838 .5893 .5774 .5837 .5558 .5780 .5802 .5881 .5624 .5772 .5810 .5819 .5779 .5850 .5829 .5838 .5780 .5273 .5475 .5281 .5792
36 .5317 .6432 .5928 .4333 .7786 .6311 .7814 .7972 .7269 .7751 .6676 .7315 .7493 .7194 .6902 .7034 .7130 .7230 .7101 .7770 .7712 .7771 .7698 .7677 .7754 .6423 .8202
39 .5106 .6704 .4590 .6326 .9370 .6650 .9689 .9681 .9529 .9666 .8329 .9648 .9622 .9582 .9161 .9439 .9569 .9624 .9523 .9663 .9625 .9634 .9615 .9306 .9602 OOM OOM
55 .3986 .8037 .6449 .7833 .8046 .7765 .8569 .8597 .8077 .8561 .8386 .8650 .8701 .8142 .8294 .8301 .8661 .8646 .8587 .8711 .8648 .8639 .8640 .8636 .8615 .8620 .8690
60 .5884 .5926 .3926 .3663 .5726 .5888 .5987 .5926 .5959 .5882 .5900 .6009 .5961 .5917 .5868 .5874 .5982 .5956 .5900 .5986 .5951 .5895 .5957 .5809 .5966 .5884 .5914
61 .2987 .4618 .4390 .4238 .6567 .4558 .6894 .6928 .6214 .6833 .5593 .6342 .6176 .5230 .5235 .4926 .6427 .6105 .6134 .6565 .6370 .6250 .6383 .5767 .6016 .5499 .7392
68 .4429 .8967 .5375 .6381 .8950 .8987 .9232 .9207 .8905 .8427 - .9301 .9463 .7264 .7736 .9338 .9377 .9207 .9154 .9335 .9283 .9343 .9397 .9121 .9262 .3689 .9650
72 .1621 .9388 .8063 .8460 .9860 .9282 .9803 .9830 .9605 .9821 .9727 .9898 .9935 .9481 .9362 .9725 .9796 .9870 .9876 .9869 .9879 .9879 .9889 .9658 .9866 .9810 .9959
88 .2500 .9625 .7125 .8275 .7450 .8348 .9090 .9357 .8760 .9098 .9677 .8672 .9227 .8953 .7852 .8807 .9250 .9115 .7978 .9367 .9373 .9550 .9542 .5723 .9477 .8450 .9575
102 .7584 .8289 .7180 .7988 .8331 .8221 .8528 .8689 .8401 .8528 .8312 .8448 .8451 .8474 .8318 .8374 .8432 .8432 .8433 .8566 .8550 .8591 .8540 .8502 .8595 .8178 .8749
103 .5945 .9614 .6188 .9236 .9233 .9342 .9768 .9752 .9756 .9764 .9653 .9728 .9733 .9728 .7779 .9725 .9726 .9725 .9710 .9703 .9502 .9683 .9662 .9196 .9665 .7652 .9742
112 .0476 .2125 .1685 .2125 .4103 .1795 .3177 .3531 .3458 .3399 - .4659 .5013 .1709 .3055 .0994 .3875 .2291 .4530 .4164 .4242 .4755 .4615 .4259 .3438 .3407 .4549
120 .3469 .6041 .5778 .5742 .5945 .5931 .6226 .6184 .6220 .6169 .6198 .6278 .6298 .5962 .5883 .6118 .6211 .6306 .6256 .6301 .6273 .6262 .6249 .6068 .6335 .6211 .6280
121 .3333 .3438 .3400 .4995 .7301 .3791 .7269 .7238 .7044 .7321 - .7372 .7352 OOM .5343 .7110 .7239 .7286 .7377 .7419 .7383 .7405 .7343 .4343 .7364 .5700 .7430
127 .9563 .9656 .8225 .1232 .9642 .9631 .9794 .9778 .9702 .9780 .9729 .9648 .9689 .9587 .9642 .9658 .9657 .9668 .9664 .9709 .9726 .9630 .9694 .9620 .9634 .9610 .9706
128 .9669 .9629 .8781 .3192 .9748 .9451 .9849 .9799 .9775 .9856 .9699 .9709 .9711 .9669 .9642 .9659 .9690 .9679 .9717 .9773 .9743 .9703 .9762 .9662 .9669 .9627 .9751
129 .3787 .9941 .9704 .9822 .9882 .9941 .9878 .9858 .9866 .9909 .9937 .9870 .9771 .9799 .9925 .9053 .9925 .9917 .9862 .9724 .9842 .9893 .9874 .9838 .9783 .9819 .9886
134 .1385 .3420 .2779 .2657 .8332 .3213 .8772 .8896 .7657 .8851 .6304 .7810 .7714 .6077 .5439 .6388 .6715 .6591 .7036 .8080 .7612 .7791 .7727 .6129 .7352 .7183 .9038
136 .4700 .5040 .4000 .5200 .5160 .5052 .6721 .6649 .5831 .6663 .5015 .5144 .5076 .5104 .4811 .4523 .5045 .5115 .4991 .6277 .5917 .5765 .5281 .5801 .5143 .4651 .5280
137 .2773 .3273 .2864 .3455 .3000 .3500 .3591 .3579 .3479 .3797 .3421 .3330 .3164 .3473 .3170 .2336 .3073 .3327 .3200 .3555 .3236 .3636 .3267 .3415 .3206 .2427 .3697
143 .9067 .9478 .8396 .8657 .9291 .9590 .9386 .9388 .9408 .9373 .9333 .9284 .9318 .9164 .9239 .9231 .9363 .9328 .9214 .9376 .9276 .9358 .9386 .9017 .9306 .9415 .9405
145 .6994 .9393 .6994 .7775 .8902 .8963 .9865 .9923 .9516 .9927 .9751 .9879 .9842 .7557 .9370 .8985 .9811 .9830 .9882 .9821 .9865 .9798 .9850 .9950 .9553 .6761 .9884
147 .4271 .4983 .5051 .4915 .5186 .5214 .5575 .5702 .5862 .5837 .5731 .5503 .5638 .5503 .5252 .4687 .5555 .5505 .5614 .5672 .5661 .5729 .5647 .4771 .5006 .4725 .5819
154 .4271 .5424 .5017 .5220 .5390 .5211 .6043 .6016 .5853 .5824 .5916 .5688 .5388 .6106 .5503 .5098 .5749 .5901 .5609 .4750 .5824 .5953 .5681 .5080 .5473 .5209 .5661
167 .5188 .9687 .9075 .9326 .8934 .9624 .9677 .9652 .9488 .9669 - .9448 .9625 .9079 .9344 .8172 .9531 .9603 .9439 .9350 .9557 .9502 .9597 .9561 .9130 OOM .9459
168 .4032 .4032 .1167 .3873 .4058 .3926 .3859 .3844 .3935 .3949 .4032 .3851 .3715 .4027 .3972 .3797 .3920 .3726 .3590 .3986 .3906 .3692 .3940 .3894 .3880 .3682 .3926
169 .2604 .9192 .8821 .8902 .9148 .9114 .9210 .9218 .9129 .9229 .9219 .9246 .9254 .9107 .8999 .9167 .9249 .9224 .9224 .9253 .9247 .9249 .9238 .9218 .9233 .3509 .9274
173 .1655 .9622 .5390 .5721 .9031 .7634 .9474 .9469 .9346 .9533 .9678 .9494 .9472 .7543 .7879 .9210 .9586 .9395 .9537 .9713 .9524 .9519 .9675 .8670 .9504 .8080 .9776
174 .3899 .4968 .4835 .4410 .5782 .4977 .7220 .7188 .6449 .7246 .5506 .6122 .6258 .6082 .5392 .5480 .6201 .5981 .5930 .7159 .6115 .6287 .6333 .5458 .6015 .4334 .9815
177 .4175 .4837 .2843 .1691 .5989 .4845 .6123 .6111 .6166 .6011 .5775 .5763 .5842 .4357 .5071 .5095 .5635 .5822 .5710 .5751 .5703 .5712 .5776 .5126 .5718 .5580 .5891
180 .2164 .9881 .5855 .5802 .9946 .9896 .9942 .9949 .9940 .9947 - .9941 .9950 .9880 .7259 .9839 .9915 .9941 .9935 .9920 .9945 .9940 .9934 .9926 .9927 .4831 .9948
196 .5743 .7979 .3674 .7678 .9310 .7954 .9298 .9308 .9311 .9309 .8159 .8012 .7680 .7976 .7973 .7356 .7982 .7974 .7984 .7973 .7994 .8002 .7987 .7412 .7975 .7618 .9309
197 .2849 .4189 .1909 .0361 .4065 .4169 .5362 .5358 .4729 .5270 - .5084 .4977 .4508 .4351 .4786 .4859 .4693 .4737 .5348 .5054 .5157 .5187 .5101 .3504 .3901 .5102
199 .5146 .6784 .6121 .6947 .8814 .6801 .8610 .8647 .8262 .8659 .8210 .9527 .8696 .8551 .7515 .8784 .8560 .8722 .9211 .9761 .9552 .9674 .9057 .7144 .9517 .6480 .9898
203 .1623 .2806 .1908 .2133 .7277 .2794 .6044 .7900 .7083 .8959 - .8606 .9006 .5547 .3841 .7894 .7808 .7952 .8570 .9138 .8879 .8889 .8922 .4274 .8446 .2435 .9156
204 .1623 .2789 .1908 .2386 .7153 .2800 .8579 .8134 .7101 .8770 - .8727 .8924 .5530 .3822 .7636 .7695 .7966 .8567 .9106 .8947 .9005 .8894 .9183 .8453 .3078 .9038
206 .1047 .7312 .7375 .7281 .7203 .7328 .7292 .7349 .7326 .7343 .7355 .7343 .7330 .7349 .7279 .7073 .7376 .7372 .7267 .7331 .7354 .7259 .7326 .7362 .7159 .7339 .7360
207 .1062 .7391 .7297 .7375 .7234 .7203 .7435 .7431 .7450 .7355 .7341 .6068 .7398 .7393 .7332 .7233 .7312 .7380 .7342 .7361 .7389 .7348 .7360 .7358 .7298 .7397 .7370
208 .0408 .7612 .5890 .6558 .9507 .6940 .9643 .9709 .9103 .9650 - .9745 .9798 .8842 .8760 .9191 .9597 .9724 .9653 .9759 .9741 .9721 .9737 .7490 .9740 .1083 .9849
211 .3990 .6355 .6010 .5764 .8276 .6256 .8213 .8128 .8204 .8332 .7750 .7241 .7629 .7392 .6158 .6640 .7015 .7097 .7291 .7123 .6801 .6677 .7166 .6683 .6939 .8105 .8325
213 .1000 .9675 .8900 .9300 .9625 .9775 .9660 .9680 .9655 .9688 - .9615 .9725 .9383 .9620 .9123 .9660 .9647 .9602 .9642 .9593 .9577 .9632 .9663 .9593 .9580 .9693
214 .1000 .8400 .7675 .7900 .8650 .8350 .8572 .8718 .8602 .8623 .8363 .8445 .8597 .8330 .8340 .7742 .8438 .8483 .8407 .8523 .8390 .8467 .8388 .8188 .8457 .8425 .8653
215 .1000 .9450 .8925 .9250 .9600 .9425 .9400 .9595 .9472 .9502 .9462 .9563 .9650 .9377 .9498 .8410 .9503 .9610 .9542 .9555 .9527 .9535 .9483 .9472 .9512 .9448 .9557
216 .1000 .7650 .7000 .6325 .7600 .7492 .7488 .7428 .7490 .7533 .7610 .7643 .7642 .6900 .6218 .7453 .7662 .7580 .7692 .7668 .7637 .7592 .7617 .7625 .7577 .7392 .7633
217 .1000 .9625 .9175 .9025 .9600 .9500 .9610 .9520 .9580 .9625 - .9623 .9655 .9455 .9572 .8685 .9592 .9642 .9588 .9490 .9585 .9600 .9560 OOM .9505 OOM .9667
218 .1000 .8225 .7050 .7300 .8050 .8238 .7807 .7855 .7603 .7945 .8158 .8477 .8362 .6793 .8257 .7263 .8385 .8397 .8347 .8333 .8317 .8298 .8250 .8172 .8332 .8103 .8727
219 .1389 .9861 .6250 .7546 1.000 .9861 .9630 .9907 .9796 .9861 .9960 .9923 .9994 .9880 .9772 .8531 .9935 .9991 .9914 .9802 .9886 .9948 .9932 .9935 .9864 .9978 .9988
220 .5580 .5717 .3468 .3370 .5863 .5680 .6203 .6189 .6112 .6192 .6060 .6300 .6293 .6116 .5964 .6142 .6284 .6190 .6220 .6276 .6218 .6290 .6222 .5967 .6263 .4639 .6328
226 .7200 .7305 .5161 .6559 .7249 .7237 .7514 .7508 .7407 .7464 .7312 .7338 .7333 .7309 .7295 .7336 .7328 .7335 .7323 .7500 .7455 .7481 .7483 .7362 .7459 .7185 .7322
227 .0094 .7719 .7438 .7281 .7781 .6531 .7715 .8533 .7804 .7640 - .8240 .8200 .7035 .7379 .1190 .7850 .8037 .7717 .8115 .7963 .8060 .8071 .5452 .7167 OOM .7898
228 .0094 .6719 .5531 .5563 .6250 .4640 .5958 .6510 .6269 .5781 - .6981 .7010 .2073 .3869 .2683 .6769 .6606 .6310 .6921 .6885 .6937 .7017 .5913 .6348 .6083 .7294
229 .0094 .8656 .7562 .6813 .8313 .8125 .8033 .8896 .7952 .8125 - .8717 .8835 .7575 .7979 .2165 .8550 .8387 .8123 .8473 .8760 .8531 .8485 .8325 .7917 OOM .8583
30
231 .1014 .9671 .9093 .7028 .9795 .9582 .9824 .9824 .9801 .9831 .9789 .9825 .9864 .9571 .9810 .9433 .9811 .9868 .9805 .9779 .9798 .9835 .9828 .9807 .9807 .9819 .9871
234 .8977 .9635 .8667 .9443 .9616 .9565 .9745 .9667 .9703 .9712 .9727 .9669 .9668 .9295 .8715 .9527 .9646 .9679 .9694 .9671 .9711 .9663 .9661 .9640 .9661 .9638 .9677
238 .1041 .9441 .8099 .8490 .9909 .9281 .9414 .9913 .9879 .9509 .9888 .9950 .9948 .9807 .9854 .9874 .9937 .9947 .9929 .9935 .9935 .9942 .9916 .9924 .9927 .2276 .9943
243 .4994 .7119 .6373 .6395 .7006 .6973 .7556 .7782 .7628 .7587 .7493 .7291 .7361 .7492 .7217 .7108 .7436 .7333 .7305 .7547 .7457 .7585 .7580 .7540 .7473 .7324 .7452
253 .2379 .8616 .7963 .8134 .9075 .8367 .9142 .9135 .9086 .9162 .9077 .8993 .9044 .8612 .8654 .8790 .8868 .8923 .8979 .8959 .8970 .9045 .9041 .8631 .8953 .8981 .9131
254 .1429 .8853 .8095 .7013 .8896 .8297 .9149 .9202 .9196 .9177 .9234 .9199 .9186 .7957 .8420 .8792 .9130 .9130 .9185 .9137 .9039 .9188 .8906 .9139 .9052 .2404 .9190
256 .1003 .9122 .8276 .7524 .9028 .8621 .9131 .9204 .9154 .9166 - .9168 .9256 .8729 .8966 .5647 .9289 .9285 .9166 .9080 .9070 .9250 .9089 .9145 .9066 .9279 .9168
260 .7859 .9657 .7556 .7147 .9991 .9268 .9998 .9997 .9997 .8539 - .9991 .9991 .9997 .9750 .9973 .9992 .9988 .9993 .9996 .9996 .9989 .9992 .9962 .9990 .9477 .9994
264 .5188 .8197 .7555 .8997 .8480 .7972 .9558 .9529 .9428 .9591 .8550 .8735 .8691 .7730 .8504 .8673 .9078 .8787 .8410 .9493 .9305 .9456 .9490 .9487 .9354 .5146 .8310
268 .3470 .7378 .5861 .6221 .7532 .7253 .7794 .7911 .7654 .7805 .7628 .7536 .7481 .5748 .4896 .7244 .7477 .7465 .7326 .7632 .7364 .6769 .7565 .7476 .7424 .2876 .7558
276 .0909 .9964 .7382 .7582 .9855 .9982 .9784 .9900 .9675 .9081 - .9964 .9976 .8820 .9800 .9863 .9968 .9985 .9953 .9962 .9972 .9967 .9950 .9943 .9932 .1962 .9981
277 .9257 .9604 .7125 .1931 .9500 .9382 .9952 .9949 .9964 .9953 .9856 .9867 .9850 .9756 .9745 .9844 .9876 .9840 .9842 .9894 .9887 .9840 .9828 .9347 .9811 .9713 .9896
278 .9245 .9682 .7086 .0874 .9563 .9571 .9915 .9952 .9929 .9974 .9823 .9842 .9808 .9353 .9727 .9743 .9814 .9830 .9814 .9858 .9847 .9807 .9824 .9825 .9759 .9612 .9861
279 .5839 .6982 .6089 .1714 .6607 .6971 .6921 .6924 .6929 .6858 .6906 .6956 .6977 .6350 .6836 .6635 .6921 .6904 .6952 .6932 .6700 .6924 .6960 .6898 .6927 .6875 .6814
282 .3050 .4802 .3273 .3591 .4777 .4357 .5103 .5170 .5090 .5147 .4994 .5026 .4980 .4751 .4660 .4851 .5023 .4934 .4988 .5153 .4980 .5070 .5128 .4692 .5139 .4750 .5102
284 .2588 .8235 .4647 .4353 .7294 .7596 .7604 .7529 .7278 .7812 .8435 .8235 .8494 .7820 .7165 .6584 .7757 .7945 .8043 .7824 .8063 .8439 .8110 .8176 .7976 .7576 .8576
285 .2196 .5875 .3852 .3108 .6741 .5742 .6974 .7088 .6271 .6960 - .6985 .7175 .6393 .6341 .6486 .6869 .6910 .6870 .7184 .6925 .7005 .7193 .6832 .6992 .5217 OOM
287 .1473 .2710 .2755 .3550 .6310 .2455 .6519 .6617 .6159 .6519 - .6434 .6445 OOM .3154 .6065 .6143 .6169 .6412 .6537 .6351 .6457 .6356 .3158 .6433 .4107 .6713
288 .4038 .7015 .5540 .5046 .9249 .6868 .9958 .9965 .9930 .9982 .9494 .9341 .9394 .9096 .8440 .9202 .9253 .9189 .9274 .9797 .9536 .9282 .9272 .9223 .9205 .9033 .9773
290 .3380 .8630 .8230 .8170 .8510 .8628 .8541 .8596 .8510 .8495 .8591 .8557 .8623 .8625 .8637 .8267 .8609 .8613 .8602 .8611 .8545 .8588 .8532 .8554 .8119 .4077 .8604
291 .3390 .3340 .3460 .3580 .3560 .3367 .3392 .3391 .3372 .3481 .3371 .3351 .3336 .3346 .3377 .3341 .3384 .3390 .3349 .3330 .3411 .3355 .3385 .3414 .3338 .3315 .3329
292 .3390 .8590 .7980 .8000 .8410 .8620 .8369 .8438 .8291 .8173 .8653 .8587 .8638 .8569 .8590 .8036 .8564 .8565 .8569 .8642 .8557 .8561 .8567 .8557 .8587 .8552 .8626
294 .5166 .8487 .7786 .8376 .8708 .8450 .8878 .9043 .8802 .8898 .8994 .8954 .8822 .8962 .8578 .8391 .8836 .8728 .8859 .9014 .8677 .9087 .8681 .8972 .8145 .8745 .8972
298 .4250 .5406 .2906 .5219 .6562 .5481 .6362 .6073 .6329 .6360 .5813 .5867 .5808 .5823 .4879 .5469 .5746 .5685 .5017 .5819 .5567 .5617 .5852 .5192 .5648 .6175 .6127
299 .4490 .5469 .3184 .4714 .6337 .5306 .6242 .6315 .6235 .6295 .5795 .5779 .5790 .5430 .5235 .4956 .5439 .5752 .5498 .5517 .6093 .5586 .5884 .5308 .5531 .3165 .6265
300 .3131 .5993 .5219 .1684 .5960 .5859 .5984 .6083 .6247 .5825 .5957 .6031 .6034 .5957 .5437 .5255 .6002 .5915 .5912 .5982 .5906 .5928 .5753 .5919 .5908 .1666 .5984
Table 15: Additional results of regression datasets. We list the average RMSE of each dataset-method
pair on 15 random seeds. ID continues the ID in Table 4. “DU” means Dummy as the naive baseline,
“LR”, “XGB”, “CB”, “RF”, “LG”, “RN”, “ND”, “ST”, “TN”, “TG”, “DAN”, “FT”, “AI”, “DCN”,
“TT”, “PT”, “GN”, “TR”, and “DR” denote “Linear Regression”, “XGBoost”, “CatBoost”, “Random
Forest”, “LightGBM”, “ResNet”, “NODE”, “SwitchTab”, “TabNet”, “TANGOS”, “DANets”, “FT-T”,
“AutoInt”, “DCNv2”, “TabTransformer”, “PTaRL”, “GrowNet”, “TabR”, and “DNNR”, respectively.
In each cell, if the experiment is failed to perform due to out of memory, we denote it by “OOM”.
The scientific notation beside the dataset name indicates the scale of the results, e.g., ×10 means the
final value of all results should be multiple by 10. And if the average RMSE of a method is too large
compared to others, i.e., still exceeds 1 after scaling, we denote it by “INF”.
1 (×104 ) .0806 .0632 .0486 .0780 .0471 .0437 .0438 .0614 .0701 .0559 .0510 .0831 .0607 .0604 .1661 .0675 .0638 .0531 .0708 .0830 .0606 .0665 .0679 .0512
2 (×101 ) .4369 .2378 .1154 .2380 .1009 .1001 .1020 .1002 .1008 .1008 .1007 .4332 .1029 .1020 .4977 .1005 .1004 .1003 .1013 .4351 .1011 .1050 .1011 .1126
3 (×100 ) .5330 .5215 .0000 .5409 .0022 .0003 .1126 .0153 .0758 .0817 .4363 .5339 .1483 .1117 .5338 .0169 .0382 .0492 .0589 .5336 .0968 .2189 .0002 .0181
4 (×101 ) .0531 .0526 .0000 .0531 .0001 .0000 .0083 .0000 .0073 .0137 .0388 .0532 .0275 .0145 .0532 .0048 .0048 .0054 .0127 .0532 .0126 .0279 .0000 .0018
6 (×101 ) .3243 .2236 .2284 .2390 .2167 .2179 .2155 .2173 .2130 .2103 .2225 .3245 .2210 .2136 .3206 .2108 .2129 .2097 .2116 .3213 .2102 .2272 .2101 .2320
7 (×10−3 ) .3993 .1679 .2047 .1696 .1527 .1465 .1559 .1503 .1563 .1572 .1519 .3832 .1641 .1564 .3841 .1554 .1553 .2719 .1627 .3981 .1554 .2377 .1522 .1660
9 (×104 ) .1851 .0756 .0746 .0762 .0749 .0817 .0751 .0747 .0774 .0793 .0760 .1826 .0862 .0786 .2326 .0774 .0799 .0787 .0794 .1872 .0762 .0840 .0791 .0829
13 (×102 ) .1570 .1179 .1158 .1188 .1149 .1143 .1141 .1143 .1173 .1167 .1150 .1564 .1177 .1175 .1555 .1153 .1161 .1171 .1177 .1784 .1166 .1185 .1142 .1260
14 (×103 ) .7282 .4797 .4620 .4855 .4560 .4557 .4617 .4555 .4631 .4607 .4556 .7180 .4606 .4626 .7285 .4547 .4565 .4583 .4594 .7253 .4572 .5434 .4594 .5123
15 (×102 ) .1044 .0623 .0490 .0632 .0460 .0458 .0458 .0457 .0467 .0463 .0462 .1026 .0465 .0471 .1099 .0460 .0465 .0468 .0471 .1097 .0465 .0721 .0460 .0495
16 (×101 ) .6588 .4482 .3112 .4536 .2922 .2902 .2988 .2953 .3036 .3007 .2909 .6467 .3075 .3044 .6399 .2930 .3025 .3015 .3062 .6526 .3006 .5483 .2933 .3460
20 (×101 ) .3038 .1511 .1082 .1518 .0889 .0761 .1053 .0912 .0764 .0717 .1161 .2982 .1128 .0785 .2978 .0702 .0801 .0724 .0910 .3010 .0798 .1334 .0573 .0837
21 (×101 ) .2416 .1033 .0786 .1038 .0648 .0537 .0798 .0647 .0601 .0531 .0802 .2407 .0888 .0600 .2496 .0540 .0593 .0554 .0681 .2407 .0625 .0943 .0455 .0678
22 (×100 ) .7817 .3121 .1300 .3245 .0696 .0337 .0507 .0538 .0265 .0551 .0358 .7519 .2721 .0601 .7373 .0174 .0215 .1289 .0384 .7711 .0706 .1914 .0343 .0970
24 (×104 ) .1635 .1432 .0623 .1530 .0453 .0487 .0467 .0465 .0512 .0513 .0550 .1791 .0575 .0506 .1776 .0560 .0537 .0515 .0539 .1637 .0515 .0642 .0515 .2925
25 (×104 ) .1635 .1455 .1347 .1552 .1199 .1275 .1228 .1256 .1408 .1372 .1401 .1748 .1426 .1316 .8299 .1382 .1716 .1330 .1502 .1768 .1326 .1469 .1458 .1270
26 (×103 ) .4368 .3797 .3671 .3842 .3654 .3646 .3692 .3646 .3659 .3665 .3642 .4395 .3664 .3658 .4362 .3654 .3671 .3650 .3672 .4378 .3639 .4019 .3664 .4488
30 (×100 ) .5007 .3301 .3293 .3448 .2749 .2638 .3052 .2769 .2344 .2183 .3040 .5014 .3230 .2148 .4986 .2768 .2453 .2360 .4859 .5103 .2272 .3086 .2427 .2478
31 (×101 ) .0501 .0326 .0315 .0346 .0269 .0259 .0286 .0271 .0223 .0231 .0303 .0510 .0324 .0210 .2297 .0262 .0236 .0217 .0235 .0517 .0213 .0319 .0226 .0270
32 (×100 ) .5007 .3409 .3324 .3564 .2738 .2674 .2945 .2795 .2433 .2292 .2315 .5614 .3638 .2237 .5464 .2839 .2641 .2312 .2544 .5407 .2381 .3284 .2368 .3306
33 (×100 ) .5007 .3371 .3389 .3633 .2829 .2721 .3046 .2868 .2701 .2568 .3080 .5202 .3270 .2465 .5876 .2815 .2807 .2519 .2740 .5122 .2514 .3166 .2433 .2868
34 (×100 ) .5007 .3460 .3371 .3602 .3098 .2954 .3255 .3105 .2609 .2540 .3271 .5456 .3833 .2355 .8969 .2958 .2767 .2645 .2759 .5167 .2456 .3444 .2796 .2837
35 (×101 ) .1536 .1514 .1491 .1532 .1483 .1486 .1488 .1483 .1504 .1520 .1497 .1567 .1500 .1498 .1594 .1520 .1544 .1511 .1621 .1546 .1496 .1592 .1549 .1715
38 (×105 ) .6554 .6357 .6117 .6327 .6012 .6000 .6011 .5992 .6548 .6528 .6549 .6288 .6194 .6557 .6108 .6548 .6543 .6771 .6553 .6042 .6133 .6077 .6506 .6862
41 (×105 ) .0398 .0166 .0076 .0151 .0053 .0052 .0055 .0053 .0123 .0083 .0059 .0392 .0112 .0099 .0405 .0053 .0054 .2439 .0054 .0404 .0067 .0304 .0051 .6178
54 (×103 ) .0293 .0321 .0240 .0284 .0178 .0159 .0169 .0170 .0250 .0374 .0188 .0296 .0257 OOM .0302 .0179 .0202 .0495 .0284 .0293 .0459 .0282 .0168 INF
57 (×102 ) .0941 .0846 .0807 .0849 .0741 .0736 .0763 .0740 .0789 .0790 .0785 .0936 .0787 .0789 .0942 .0787 .0789 .0789 .0790 .0941 .0787 .0804 .0742 INF
62 (×100 ) .4890 .4845 .4687 .4878 .4320 .4320 .4287 .4247 .4844 .4931 .4840 .4898 .5916 .4846 .4889 .4843 .4927 .6600 .4985 .4904 .4845 .4914 .4839 .7022
64 (×100 ) .4977 .3509 .2514 .3584 .2628 .2609 .2720 .2758 .3121 .3245 .3014 .5154 .3493 .2889 .6651 .2933 .3012 .3307 .3169 .5139 .2924 .3496 .3180 .3047
66 (×100 ) .3131 .0568 .0837 .0589 .0406 .0348 .0448 .0420 .0284 .0265 .0502 .3200 .0571 .0290 .3396 .0296 .0314 .0261 .0346 .3254 .0325 .0777 .0199 .2158
69 (×100 ) .4286 .4289 .4303 .4299 .4286 .4288 .4301 .4286 .4592 .4376 .4287 .4337 .4296 .4326 .4480 .4397 .4392 .4827 .4569 .4421 .4286 .4461 .4331 .4847
70 (×100 ) .5004 .2798 .2345 .2796 .2218 .2135 .2293 .2198 .2369 .2194 .2055 .5008 .2618 .2191 .7146 .2137 .2148 .2327 .2071 .5031 .2030 .2404 .2070 .2325
71 (×100 ) .3202 .3124 .3088 .3203 .3075 .3065 .3035 .3060 .3280 .3159 .3120 .3299 .3203 .3215 .3280 .3394 .3269 .3294 .3345 .3304 .3138 .3321 .3237 .3401
73 (×102 ) .1330 .1322 .1321 .1326 .1053 .1024 .0766 .0768 .1325 .1322 .1323 .1331 .1320 .1326 .1330 .1321 .1324 .1321 .1326 .1330 .1322 .1325 .1143 .2095
74 (×101 ) .1000 .0817 .0831 .0827 .0705 .0707 .0717 .0708 .0906 .0811 .0737 .1036 .0734 .0812 .1140 .0716 .0781 .0834 .0810 .1044 .0790 .1003 .0906 .2550
76 (×103 ) .1817 .1515 .1116 .1561 .0718 .0726 .0760 .0716 .0781 .0764 .0783 .1818 .0776 .0787 .1864 .0763 .0762 .0723 .0735 .1854 .0714 .0845 .0735 .0850
77 (×103 ) .8683 .5908 .4968 .6939 .4448 .4506 .4587 .4528 .4789 .4736 .5112 .8839 .5106 .4641 .8800 .4609 .4714 .4652 .4610 .8703 .4777 .4996 .4606 .5419
78 (×106 ) .2031 .0269 .0803 .0303 .0182 .0187 .0172 .0180 .0221 .0215 .0260 .2086 .0289 .0230 .3268 .0192 .0230 .0215 .0280 .2034 .0225 .0625 .0232 INF
31
79 (×105 ) INF .1319 .3229 .1384 .0692 .0717 .0729 .0796 .0820 .0884 .0757 INF .1101 .0839 INF .0772 .0842 .0796 .1071 INF .0850 .1666 .0705 INF
80 (×107 ) .0105 .0066 .0043 .0033 .0018 .0019 .0018 .0019 .0023 .0025 .0030 .0105 .0039 .0027 .0105 .0019 .0028 .0033 .0028 .0105 .0024 .0105 .0022 .5791
81 (×105 ) .7019 .1276 .1176 .1465 .0492 .0450 .0629 .0493 .0600 .0598 .0543 .7202 .1059 .0590 INF .0437 .0562 .0741 .0701 .6984 .0584 .1295 .0389 INF
82 (×101 ) .0453 .0322 .0282 .0330 .0091 .0122 .0091 .0077 .0123 .0115 .0295 .0472 .0104 .0130 .1242 .0088 .0109 .0133 .0103 .0455 .0113 .0227 .0101 .0219
86 (×106 ) .3138 .1692 .1014 .2050 .0874 .0803 .0980 .0881 .0901 .0913 .0958 .3184 .1055 .0922 .3174 .0881 .0883 .0882 .0946 .3150 .0953 .1190 .0872 .0892
87 (×104 ) .2939 .2163 .1113 .2312 .0703 .0775 .0775 .0797 .1112 .1115 .1372 .2958 .1561 .1107 .2959 .0809 .0760 .0820 .0772 .2996 .0795 .1098 .0889 .1093
89 (×102 ) .6831 .4395 .4141 .4488 .4069 .4037 .4097 .4052 .4066 .4045 .4028 .7061 .4069 .4064 .6632 .4028 .4050 .4052 .4073 .6805 .4053 .4090 .4031 .4715
90 (×102 ) .1963 .1741 .1608 .1750 .1559 .1582 .1560 .1564 .1552 .1546 .1531 .1981 .1688 .1549 .1963 .1530 .1539 .1587 .1577 .1970 .1553 .1725 .1563 .1778
92 (×102 ) .1679 .1564 .1586 .1644 .1557 .1548 .1561 .1556 .1724 .1644 .1606 .1856 .1683 .1545 .1948 .1782 .1763 .1546 .1774 .1817 .1649 .1735 .1825 .1526
93 (×101 ) .8339 .7709 .4835 .7772 .1515 .1599 .1710 .1502 .8308 .8353 .8333 .5872 .3690 .8333 .3286 .8344 .8323 .8324 .8203 .4317 .3057 .5429 .8003 .3953
94 (×100 ) .4830 .4270 .4271 .4264 .4294 .4425 .4302 .4334 .4691 .4703 .4465 .5081 .4549 .4806 .6890 .4388 .4606 .4516 .4540 .5325 .4285 .4436 .4439 .4521
96 (×101 ) .6120 .5141 .3690 .5237 .3661 .3475 .3911 .3744 .3675 .3446 .4230 .6064 .3939 .3530 .6248 .3649 .3641 .3549 .4053 .6130 .3474 .5679 .3047 .3765
105 (×100 ) .4921 .4446 .4428 .4551 .4262 .4266 .4292 .4261 .5028 .4416 .4296 .4979 .4303 .4521 .5468 .4336 .4407 .4544 .4340 .5101 .4298 .4447 .5212 .4750
106 (×102 ) .2846 .2867 .2854 .2859 .2863 .2864 .2936 .2848 .2874 .2926 .2873 .2889 .2970 .2869 .2923 .2868 .2883 .2884 .2890 .2853 .2862 .2859 .2875 .3267
108 (×101 ) .4429 .4138 .4112 .4174 .4047 .4023 .4087 .4022 .4454 .4448 .4470 .4280 .4898 .4494 .4402 .4492 .4474 .4613 .4645 .5102 .4124 .4245 .4468 .4353
109 (×101 ) .4735 .1823 .3060 .1572 .1535 .1849 .1270 .1409 .1996 .1989 .1648 .4675 .3291 .1859 .4692 .1488 .1836 .2136 .2288 .4725 .2009 .4089 .1929 .3169
110 (×103 ) .0343 .0180 .0108 .0184 .0100 .0099 .0106 .0102 .0104 .0105 .0127 .0336 .0128 .0109 .0371 .0106 .0108 .0103 .0113 .0342 .0108 .0148 .0103 INF
113 (×103 ) .1234 INF .0090 .0312 .0105 .0091 .0232 .0098 .0100 .0116 .0086 .1202 .0154 .0097 .1259 .0126 OOM .0193 .0494 .1222 .0104 .0498 OOM INF
114 (×100 ) .1021 .0991 .0997 .1059 .1010 .1044 .1023 .1014 .1011 .1022 .1011 .1025 .1012 .1023 .1026 .1023 .1020 .1041 .1046 .1025 .0990 .1015 .1024 .1107
116 (×100 ) .5005 .4489 .4519 .5822 .4589 .4705 .4565 .4793 .4595 .4586 .4758 .5254 .4911 .4546 .9514 .4563 .4496 .4681 .4636 .5425 .4413 .4647 .5101 .6206
118 (×100 ) .7814 .6702 .6532 .6900 .6192 .6081 .6183 .6193 .6693 .6601 .6512 .8179 .7090 .6644 .8191 .6514 .6698 .6754 .6639 .7875 .6598 .6739 .6745 .7198
119 (×100 ) .8972 .7700 .6277 .7605 .6235 .6132 .6466 .6316 .6925 .6644 .6924 .9103 .7466 .6934 .9056 .7003 .6757 .7280 .6689 .9031 .6933 .7005 .6632 .6944
125 (×102 ) .0708 .0474 .0272 .0486 .0177 .0139 .0201 .0178 .0154 .0130 .0130 .0709 .0236 .0156 .1386 .0126 .0142 .0133 .0197 .0709 .0156 .0354 .0126 .0235
130 (×100 ) .5135 .3996 .1097 .4616 .0808 .0764 .0865 .0852 .0954 .0986 .1014 .5204 .1533 .1072 .5365 .0916 .0976 .0943 .1055 .5145 .0958 .1553 .0897 .1146
131 (×103 ) .4612 .4122 .3557 .4156 .3332 .3336 .3454 .3393 .3705 .3571 .3448 .4598 .4307 .3739 .4632 .3907 .3730 .3986 .3598 .4653 .3611 .4067 .3969 .3938
132 (×101 ) .4430 .4394 .4345 .4390 .4350 .4348 .4499 .4443 .4326 .4478 .4182 .4442 .5458 .4290 .4712 .4351 .4413 .4372 .4452 .4466 .4449 .4316 .4636 .4508
133 (×101 ) .3454 .2902 .3054 .2955 .2980 .2924 .2739 .2762 .2938 .2972 .2874 .3455 .3841 .2943 .3434 .2964 .2920 .2957 .2934 .7863 .2982 .3044 .3075 .2938
135 (×105 ) .0986 .0769 .0459 .0839 .0068 .0079 .0104 .0053 .1016 .0963 .0983 .0520 .0355 .0968 .0273 .1015 .1021 .1012 .0959 .0274 .0302 .0511 .0937 .0296
138 (×100 ) .3974 .2944 .2405 .2956 .1148 .1028 .1478 .1172 .1788 .1718 .2356 .3885 .2332 .2026 .3963 .1100 .1221 .1253 .1352 .3952 .1358 .3467 .0908 .1893
140 (×100 ) .1202 .0815 .0969 .0860 .0810 .0792 .0833 .0810 .0789 .0794 .0786 .1241 .0848 .0786 .1330 .0791 .0797 .0786 .0813 .1238 .0778 .1234 .0782 .0961
141 (×100 ) .1551 .0408 .0493 .0409 .0301 .0288 .0319 .0303 .0297 .0289 OOM .1549 .0313 .0294 .1588 .0286 .0291 .0288 .0291 .1544 .0285 .0388 .0293 .0307
144 (×103 ) .0343 INF .0226 .0303 .0208 .0204 .0202 .0204 .0238 .0247 .0229 .0348 .0263 .0238 .0553 .0232 OOM .0240 .0284 OOM .0233 .0333 .0224 INF
148 (×102 ) .1707 .0463 .0393 .0463 .0344 .0314 .0361 .0327 .1603 .1540 .1698 .0674 .0410 .1651 .0405 .1694 .1653 .1476 .1252 .0737 .0395 .0499 .1627 .0396
149 (×100 ) .2213 .1375 .1348 .1380 .1325 .1336 .1337 .1363 .1338 .1425 .1363 .2294 .1530 .1355 .3889 .1369 .1386 .1361 .1366 .2218 .1414 .1449 .1437 .1478
152 (×101 ) .0500 .0455 .0415 .0461 .0392 .0390 .0425 .0409 .0479 .0452 .0442 .0525 .0447 .0433 .1065 .0447 .0456 .0471 .0449 .0500 .0433 .0513 .0355 .2482
153 (×102 ) .1630 .1063 .0935 .1060 .0483 .0438 .0644 .0466 .1609 .1559 .1624 .1007 .0751 .1630 .0614 .1626 .1598 .1619 .1546 .0682 .0515 .0696 .1573 .0712
155 (×102 ) .1735 .0892 .0478 .1011 .0240 .0236 .0253 .0230 .0249 .0248 .0250 .1711 .0322 .0254 .1746 .0233 .0248 .0247 .0270 .1731 .0249 .0439 .0233 .0256
156 (×102 ) .1735 .0930 .0364 .1021 .0278 .0271 .0293 .0274 .0291 .0298 .0306 .1727 .0343 .0294 .1742 .0285 .0291 .0294 .0313 .1739 .0292 .0475 .0278 .0315
158 (×101 ) .0500 .0472 .0457 .0485 .0392 .0389 .0392 .0391 .0423 .0413 .0421 .0501 .0424 .0415 .0520 .0410 .0402 .0415 .0434 .0505 .0396 .0432 .0434 .4232
161 (×101 ) .6419 .4430 .4393 .4426 .3967 .3959 .4070 .4044 .3977 .3971 .4156 .6487 .4048 .4012 .6509 .4081 .4186 .4043 .4140 .6441 .3931 .4115 .4243 .4337
162 (×100 ) .1485 .1370 .0663 .1393 .0720 .0643 .0830 .0641 .0650 .0650 .1192 .1501 .1080 .0695 .1492 .0636 .0687 .0655 .0706 .1484 .0678 .1023 .0724 .0776
165 (×10−2 ) .2342 .1416 .1410 .1423 .1388 .1383 .1399 .1395 .1381 .1389 .1371 .2290 .1406 .1375 .2338 .1378 .1371 .1389 .1384 .2379 .1379 .1401 .1386 .1598
172 (×10−2 ) .6519 .2812 .3703 .2899 .2340 .2030 .2815 .2310 .1868 .1885 .2230 .6556 .1981 .1898 .6717 .1827 .1874 .1850 .1896 .6588 .1867 .2328 .1844 .2195
176 (×101 ) .1346 .1088 .0937 .1091 .0772 .0773 .0787 .0770 .0814 .0814 .0796 .1311 .0809 .0815 .1402 .0798 .0800 .0807 .0826 .1340 .0804 .0836 .0788 .0885
178 (×101 ) .4958 .2628 .1854 .2643 .1079 .1010 .1395 .1063 .1022 .1029 .1007 .4945 .1032 .1041 .4949 .1009 .1018 .1010 .1048 .4935 .1026 .1144 .1017 .1119
179 (×100 ) .1778 .1510 .1441 .1570 .1281 .1315 .1299 .1281 .1821 .1756 .1761 .1534 .1590 .1779 .1376 .1763 .1757 .1765 .1715 .1649 .1542 .1589 .1758 .1506
181 (×102 ) .1564 .0095 .0101 .0095 .0057 .0052 .0073 .0059 .1539 .1403 .1554 .0216 .0078 .1562 .0065 .1548 .1521 .1633 .1338 .0145 .0056 .0144 .1539 .0048
184 (×105 ) .1250 .0615 .0532 .0651 .0465 .0474 .0453 .0470 .0480 .0485 .0468 .1283 .0564 .0487 .1310 .0462 .0472 .0487 .0522 .1248 .0476 .1116 .0465 .0530
188 (×105 ) .5300 .4571 .3623 .4819 .3136 .2994 .3278 .3098 .3175 .3137 .3483 .5397 .3533 .3187 .5458 .3067 .3123 .3133 .3280 .5404 .3158 .5156 .3054 .4125
189 (×106 ) .0530 .0413 .0315 .0480 .0287 .0285 .0289 .0285 .0297 .0293 .0300 .0535 .0312 .0298 .1004 .0291 .0294 .0298 .0310 .0532 .0297 .0311 .0292 .0361
190 (×106 ) .3890 .2130 .1714 .2406 .1396 .1177 .1485 .1359 .1382 .1312 .1376 .3928 .1526 .1413 .4149 .1193 .1279 .1309 .1453 .3931 .1394 .1778 .1423 .1332
191 (×106 ) .1145 .0701 .0600 .0714 .0480 .0455 .0525 .0466 .0520 .0504 .0543 .1148 .0542 .0520 .1167 .0480 .0521 .0502 .0518 .1149 .0513 .0580 .0421 .0563
192 (×107 ) .1622 .1111 .1121 .1120 .1020 .1117 .1141 .1027 .1113 .1120 .1045 .1598 .1384 .1138 .1510 .1081 .1069 .1176 .1204 .2317 .1148 .1263 .1108 .1173
202 (×100 ) .2631 .2027 .1205 .2054 .1322 .0902 .1492 .1215 .0688 .0670 .0935 .2620 .0830 .0716 .4072 .0674 .0699 .0668 .0743 .2619 .0678 .1111 .0664 .0973
205 (×10−1 ) INF INF INF INF .1453 .0002 .0000 .0000 .0751 .0645 .0064 INF .1157 .0414 INF .0024 .0287 .0064 .0065 INF .0205 INF .0023 .4997
212 (×102 ) .1724 .0230 .0249 .0235 .0059 .0041 .0066 .0044 .0064 .0091 .0042 .1672 .0157 .0063 .2268 .0055 .0065 .0057 .0068 .1714 .0067 .0345 .0044 .0118
223 (×101 ) INF .4474 .1478 .4975 .0193 .0085 .0091 .0114 .0029 .0139 .0009 INF .0208 .0238 INF .0031 .0050 .0016 .0073 INF .0097 .0719 .0028 .0034
233 (×100 ) .1326 .1383 .1416 .1317 .1423 .1397 .1350 .1412 .1393 .1589 .1321 .1332 .1445 .1369 .1329 .1330 .1376 .1395 .1502 .1328 .1464 .1333 .1341 .1458
241 (×102 ) .4126 .3044 .0892 .3057 .0478 .0419 .0542 .0504 .0244 .0309 .0442 .4170 .0433 .0253 .4051 .0216 .0231 .0221 .0314 .4072 .0252 .3970 .0224 .0965
242 (×102 ) .4155 .3044 .0933 .3095 .0438 .0367 .0498 .0458 .0258 .0310 .0442 .4170 .0440 .0256 .5338 .0222 .0251 .0231 .0330 .4275 .0455 .0993 .0232 .0995
244 (×10−1 ) .2943 .2601 .2567 .2608 .0787 .0645 .0791 .0780 .0817 .0639 .0863 .2939 .0965 .0897 .2941 .0754 .0676 .0653 .0757 .2941 .0701 .2544 .0591 .0918
245 (×101 ) .5571 .4435 .3768 .4440 .3304 .3281 .3272 .3279 .3341 .3373 OOM .5517 .3328 .3366 .5563 .3285 .3318 .3321 .3366 .5552 .3300 .3631 .3308 .3640
247 (×101 ) .1727 .1254 .1112 .1427 .1136 .1062 .1154 .1244 .1162 .1193 .1122 .1711 .1298 .1169 .1731 .1239 .1143 .1172 .1180 .1714 .1126 .1166 .1172 .1247
248 (×101 ) .1259 .0867 .0964 .0982 .0903 .0862 .0860 .0879 .0889 .0914 .0857 .1256 .0957 .0880 .1310 .0907 .0908 .1029 .0876 .1261 .0947 .1123 .0948 .0989
252 (×101 ) .2205 .1207 .0711 .1227 .0703 .0668 .0722 .0690 .0689 .0739 .0834 .2170 .0838 .0685 .3361 .0727 .0726 .0699 .0755 .2157 .0716 .2118 .0727 .0784
257 (×100 ) .8345 .8358 .7962 .8879 .7245 .7331 .7073 .7037 .7747 .8095 .7300 .8328 .8709 .8148 .8312 .7234 .7554 .7457 .7566 .8313 .7344 .9048 .7812 .7968
258 (×100 ) .3057 .2771 .2641 .3041 .2658 .2637 .2707 .2639 .2658 .2631 .2622 .3109 .2606 .2616 .3244 .2760 .2765 .2642 .2729 .3186 .2608 .2693 .2974 .2958
259 (×100 ) .3956 .3668 .3344 .3882 .3161 .3127 .3188 .3150 .3526 .3291 .3173 .4023 .3259 .3233 .3980 .3222 .3308 .3452 .3392 .4061 .3197 .3555 .3185 .3553
261 (×103 ) .0412 .0216 .0236 .0194 .0239 .0246 .0250 .0234 .0280 .0289 .0275 .0415 .0276 .0210 .9859 .0135 .0146 .0223 .0129 .0419 .0111 .0375 .0198 .0278
262 (×100 ) .2036 .1387 .1269 .1400 .1138 .1091 .1191 .1109 .0978 .0983 .1319 .1983 .1374 .1028 .2037 .1017 .1025 .1057 .1004 .2002 .0989 .1174 .0994 .1046
267 (×102 ) .3333 .0483 .0295 .0496 .0108 .0076 .0135 .0107 .0074 .0102 .0037 .3325 .0128 .0163 .3541 .0051 .0067 .0054 .0076 .3478 .0086 .0374 .0040 .0149
269 (×102 ) .0688 .0212 .0073 .0213 .0084 .0080 .0097 .0091 .0076 .0074 .0070 .0673 .0098 .0077 .1182 .0076 .0075 .0082 .0077 .0680 .0091 .0158 .0069 .0080
270 (×102 ) .1980 .2090 .1911 .2187 .2236 .1835 .1802 .1843 .1986 .1954 .1868 .2000 .2451 .2058 .1996 .1887 .1915 .1873 .2099 .1990 .2180 .1915 .1694 .2198
271 (×10−1 ) .5664 .4659 .2065 .4945 .2185 .2166 .2471 .2438 .2102 .1871 .3405 .5669 .2893 .2164 .5677 .2127 .1888 .2234 .1858 .5658 .2371 .3174 .1794 .2221
272 (×100 ) .8405 .7736 .7717 .8137 .6877 .6704 .6827 .6829 .7739 .7409 .7066 .8920 .7815 .6994 .9334 .7379 .7581 .7452 .8161 .9150 .7042 .7506 .7410 .7567
280 (×100 ) .0356 .0389 .0352 .0353 .0348 .0347 .0348 .0348 .0351 .0353 .0350 .0359 .0354 .0352 .0358 .0355 OOM .0694 .0359 .0357 .0347 .0364 .0358 INF
281 (×102 ) .0339 .0028 .0030 .0029 .0029 .0028 .0032 .0031 .0021 .0022 .0033 .0424 .0055 .0021 .1571 .0031 .0025 .0022 .0026 .0319 .0024 .0068 .0028 .0023
286 (×103 ) .0610 .0556 .0516 .0563 .0380 .0382 .0390 .0385 .0480 .0558 .0429 .0611 .0488 .0470 .0625 .0421 .0462 .0510 .0480 .0613 .0475 .0578 .0416 INF
293 (×102 ) .1356 .0132 .0201 .0133 .0125 .0118 .0136 .0133 .0117 .0212 .1040 .1303 .0169 .0121 .1438 .0117 .0120 .0114 .0149 .1358 .0121 .0339 .0113 .0220
295 (×102 ) .0669 .0316 .0335 .0316 .0300 .0292 .0308 .0301 .0288 .0290 .0348 .0635 .0318 .0293 .1472 .0294 .0298 .0292 .0294 .0663 .0290 .0317 .0291 .0352
297 (×100 ) .9022 .7666 .6793 .7690 .6613 .6569 .6758 .6751 .7314 .7105 OOM .9149 .7483 .7470 .9099 .7287 .7244 .7441 .7243 .9012 .7192 .7391 .6895 .7767
32
Table 16: Additional results of binary classification datasets. For each dataset-method pair, we list
the standard deviation of accuracy on 15 random seeds. ID continues the ID in Table 4. “DU”
means Dummy as the naive baseline, “LR”, “NB”, “XGB”, “CB”, “RF”, “LG”, “PFN”, “RN”, “ND”,
“ST”, “TN”, “TC”, “TG”, “DAN”, “FT”, “AI”, “DCN”, “TT”, “PT”, “GN”, and “TR” denote “Logistic
Regression”, “Naive Bayes”, “XGBoost”, “CatBoost”, “Random Forest”, “LightGBM”, “TabPFN”,
“ResNet”, “NODE”, “SwitchTab”, “TabNet”, “TabCaps”, “TANGOS”, “DANets”, “FT-T”, “AutoInt”,
“DCNv2”, “TabTransformer”, “PTaRL”, “GrowNet”, and “TabR”, respectively. In each cell, if the
experiment is failed to perform due to out of memory, we denote it by “OOM”. TabPFN cannot
handle datasets with more than 10 classes or more than 100 features.
8 .0000 .0000 .0000 .0000 .0000 .0000 .0009 .0006 .0006 .0008 .0000 .0010 .0008 .0000 .0004 .0004 .0001 .0008 .0010 .0012 .0014 .0008 .0011 .0008 .0009 .0046 .0008
11 .0000 .0000 .0000 .0000 .0000 .0001 .0006 .0005 .0005 .0003 .0010 .0005 .0005 .0007 .0008 .0011 .0007 .0005 .0005 .0005 .0007 .0007 .0005 .0016 .0005 .0006 .0007
17 .0000 .0000 .0000 .0000 .0000 .0000 .0014 .0020 .0009 .0000 .0046 .0017 .0014 .0008 .0030 .0058 .0032 .0019 .0033 .0012 .0017 .0014 .0015 .0013 .0013 .0019 .0023
18 .0000 .0000 .0000 .0000 .0000 .0046 .0016 .0011 .0020 .0018 .0025 .0030 .0017 .0008 .0043 .0051 .0039 .0035 .0033 .0030 .0036 .0023 .0032 .0053 .0038 .0037 .0026
19 .0000 .0000 .0000 .0000 .0000 .0044 .0114 .0156 .0114 .0057 .0063 .0174 .0110 .0212 .0185 .0191 .0090 .0071 .0120 .0082 .0133 .0135 .0134 .0114 .0083 .0292 .0092
23 .0000 .0000 .0000 .0000 .0000 .0010 .0003 .0003 .0003 .0002 .0010 .0003 .0003 .0002 .0002 .0007 .0003 .0004 .0003 .0004 .0003 .0002 .0002 .0003 .0003 .0005 .0002
27 .0000 .0000 .0000 .0000 .0000 .0005 .0022 .0013 .0014 .0024 .0025 .0036 .0030 .0021 .0035 .0081 .0057 .0044 .0021 .0019 .0023 .0020 .0038 .0027 .0017 .0026 .0024
28 .0000 .0000 .0000 .0000 .0000 .0200 .0009 .0009 .0005 .0007 .0020 .0011 .0008 .0020 .0012 .0087 .0015 .0010 .0010 .0013 .0012 .0014 .0014 .0021 .0010 .0038 .0006
29 .0000 .0000 .0000 .0000 .0000 .0717 .0003 .0002 .0002 .0000 .0000 .0002 .0001 .0001 .0004 .0002 .0002 .0001 .0007 .0003 .0004 .0005 .0005 .0005 .0004 .0006 .0004
37 .0000 .0000 .0000 .0000 .0000 .0055 .0070 .0022 .0043 .0041 .0031 .0116 .0152 .0042 .0071 .0095 .0042 .0127 .0137 .0053 .0092 .0041 .0080 .0059 .0090 .0039 .0072
40 .0000 .0000 .0000 .0000 .0000 .0058 .0131 .0145 .0153 .0112 .0073 .0145 .0139 .0133 .0350 .0548 .0194 .0116 .0169 .0193 .0231 .0105 .0172 .0187 .0263 .0352 .0147
42 .0000 .0000 .0000 .0000 .0000 .0000 .0092 .0062 .0047 .0018 .0043 .0031 .0020 .0014 .0064 .0085 .0082 .0050 .0039 .0034 .0058 .0042 .0101 .0058 .0037 .0068 .0037
43 .0000 .0000 .0000 .0000 .0000 .0082 .0030 .0040 .0022 .0028 .0023 .0045 .0029 .0056 .0190 .0144 .0054 .0028 .0036 .0029 .0038 .0023 .0026 .0048 .0326 .0035 .0034
44 .0000 .0000 .0000 .0000 .0000 .0000 .0032 .0044 .0029 .0023 .0027 .0036 .0036 .0026 .0044 .0073 .0031 .0034 .0030 .0026 .0036 .0027 .0034 .0020 .0032 .0033 .0039
45 .0000 .0000 .0000 .0000 .0000 .0105 .0116 .0115 .0107 .0086 .0045 .0101 .0090 .0055 .0227 .0520 .0162 .0109 .0480 .0142 .0440 .0202 .0286 .0100 .0145 .0222 .0066
46 .0000 .0000 .0000 .0000 .0000 .0044 .0035 .0020 .0043 .0033 .0060 .0018 .0023 .0024 .0357 .0022 .0055 .0801 .0032 .0147 .0043 .0026 .0035 .0030 .0018 .0051 .0016
47 .0000 .0000 .0000 .0000 .0000 .0000 .0122 .0145 .0113 .0107 .0088 .0890 .0140 .1229 .0395 .0301 .1164 .0211 .0859 .0181 .0809 .0123 .0280 .0147 .1171 .0268 .0108
48 .0000 .0000 .0000 .0000 .0000 .0016 .0132 .0085 .0107 .0157 .0086 .0209 .0129 .0112 .0470 .0665 .0496 .0238 .0765 .0741 .0478 .0206 .0219 .0599 .0245 .0276 .0255
49 .0000 .0000 .0000 .0000 .0000 .0108 .0031 .0038 .0026 .0031 .0073 .0022 .0033 .0033 .0227 .0782 .0076 .0043 .0074 .0061 .0093 .0076 .0026 .0042 .0089 .0160 .0020
50 .0000 .0000 .0000 .0000 .0000 .0012 .0028 .0029 .0027 .0040 .0406 .0049 .0021 .0005 .0077 .0842 .0638 .0022 .0644 .0649 .0073 .0120 .0061 .0280 .0781 .0128 .0014
51 .0000 .0000 .0000 .0000 .0000 .0171 .0028 .0023 .0024 .0035 .0159 .0406 .0412 .0039 .0032 .0582 .0514 .0038 .0511 .0379 .0461 .0079 .0056 .0047 .0074 .0161 .0018
52 .0000 .0000 .0000 .0000 .0000 .0000 .0149 .0214 .0095 .0000 .0054 .0196 .0187 .0167 .0272 .0433 .0248 .0150 .0348 .0579 .0199 .0123 .0776 .0094 .0978 .0115 .0136
53 .0000 .0000 .0000 .0000 .0000 .0002 .0056 .0016 .0027 .0024 .0378 .0041 .0035 .0079 .0066 .0906 .0144 .0824 .0063 .0052 .0078 .0079 .0075 .0042 .0055 .0136 .0780
56 .0000 .0000 .0000 .0000 .0000 .0000 .0075 .0143 .0056 .0072 .0050 .0121 .0246 .0056 .0153 .0277 .0069 .0036 .0104 .0291 .0198 .0098 .0127 .0148 .0098 .0281 .0070
58 .0000 .0000 .0000 .0000 .0000 .0000 .0132 .0079 .0167 .0189 .0288 .0200 .0240 .0750 .0340 .0210 .0271 .0172 .0218 .0232 .0681 .0820 .0362 .0217 .0421 .0361 .0491
59 .0000 .0000 .0000 .0000 .0000 .0278 .0110 .0057 .0092 .0068 .0133 .0158 .0164 .0663 .0409 .0258 .0101 .0163 .0146 .0745 .0767 .0114 .0134 .0124 .0753 .0109 .0109
63 .0000 .0000 .0000 .0000 .0000 .0063 .0021 .0008 .0020 .0000 .0022 .0014 .0022 .0012 .0032 .0065 .0046 .0021 .0031 .0021 .0027 .0016 .0027 .0031 .0013 .0053 .0034
65 .0000 .0000 .0000 .0000 .0000 .0015 .0056 .0045 .0037 .0076 .0025 .0044 .0077 .0065 .0061 .0058 .0101 .0054 .0033 .0081 .0081 .0095 .0083 .0121 .0107 .0113 .0080
67 .0000 .0000 .0000 .0000 .0000 .0027 .0013 .0014 .0008 .0011 .0041 .0022 .0041 .0009 .0028 .0048 .0034 .0026 .0023 .0022 .0022 .0028 .0037 .0024 .0030 .0039 .0016
75 .0000 .0000 .0000 .0000 .0000 .0000 .0023 .0038 .0041 .0016 .0043 .0055 .0070 .0027 .0067 .0322 .0077 .0065 .0071 .0065 .0068 .0058 .0075 .0108 .0042 .0083 .0086
83 .0000 .0000 .0000 .0000 .0000 .0000 .0062 .0098 .0031 .0052 - .0080 .0090 .0074 .0169 .0074 .0116 .0073 .0101 .0117 .0073 .0117 .0097 .0147 .0125 .0132 .0176
84 .0000 .0000 .0000 .0000 .0000 .0000 .0017 .0017 .0017 .0000 .0025 .0024 .0026 .0007 .0052 .0019 .0021 .0020 .0021 .0021 .0023 .0036 .0021 .0034 .0021 .0027 .0021
85 .0000 .0000 .0000 .0000 .0000 .0065 .0030 .0048 .0040 .0061 .0025 .0081 .0060 .0083 .0076 .0108 .0097 .0055 .0068 .0050 .0064 .0040 .0063 .0095 .0098 .0059 .0070
91 .0000 .0000 .0000 .0000 .0000 .0045 .0046 .0073 .0033 .0035 .0012 .0070 .0064 .0023 .0055 .0080 .0055 .0072 .0062 .0026 .0036 .0051 .0056 .0079 .0052 .0088 .0073
95 .0000 .0000 .0000 .0000 .0000 .0000 .0013 .0012 .0010 .0011 .0026 .0016 .0023 .0052 .0035 .0096 .0018 .0020 .0020 .0024 .0024 .0014 .0015 .0013 .0028 .0026 .0007
97 .0000 .0000 .0000 .0000 .0000 .0000 .0052 .0000 .0033 .0059 .0029 .0044 .0080 .0012 .0135 .1845 .0061 .0037 .0057 .0117 .0062 .0189 .0098 .0037 .0012 .0141 .0112
98 .0000 .0000 .0000 .0000 .0000 .0000 .0110 .0084 .0132 .0120 .0057 .0118 .0136 .0066 .0239 .0222 .0191 .0102 .0301 .0106 .0173 .0185 .0132 .0162 .0107 .0137 .0125
99 .0000 .0000 .0000 .0000 .0000 .0029 .0079 .0074 .0037 .0027 .0019 .0071 .0060 .0012 .0102 .0073 .0068 .0026 .0083 .0060 .0066 .0092 .0107 .0080 .0042 .0069 .0067
100 .0000 .0000 .0000 .0000 .0000 .0273 .0032 .0054 .0023 .0024 .0037 .0057 .0030 .0027 .0128 .0110 .0044 .0030 .0220 .0048 .0035 .0044 .0045 .0076 .0051 .0064 .0062
101 .0000 .0000 .0000 .0000 .0000 .0057 .0091 .0130 .0085 .0059 .0022 .0155 .0126 .0554 .0192 .0625 .0157 .0106 .0120 .0096 .0293 .0120 .0173 .0203 .0115 .0126 .0156
104 .0000 .0000 .0000 .0000 .0000 .0000 .0002 .0003 .0003 .0017 .0011 .0007 .0011 .0017 .0005 .0014 .0006 .0003 .0004 .0005 .0017 .0010 .0014 .0013 .0010 .0015 .0005
107 .0000 .0000 .0000 .0000 .0000 .0012 .0002 .0002 .0003 .0004 .0029 .0003 .0003 .0002 .0008 .0005 .0004 .0003 .0003 .0003 .0004 .0002 .0003 .0004 .0005 .0004 OOM
111 .0000 .0000 .0000 .0000 .0000 .0000 .0029 .0018 .0041 .0022 .0024 .0061 .0070 .0048 .0096 .0159 .0180 .0064 .0077 .0079 .0077 .0062 .0065 .0073 .0050 .0086 .0053
115 .0000 .0000 .0000 .0000 .0000 .0000 .0059 .0111 .0068 .0070 .0035 .0124 .0115 .0158 .0091 .0121 .0089 .0144 .0174 .0139 .0129 .0108 .0120 .0127 .0118 .0111 .0166
117 .0000 .0000 .0000 .0000 .0000 .0018 .0016 .0010 .0010 .0011 .0005 .0024 .0046 .0000 .0054 .0138 .0039 .0041 .0054 .0039 .0021 .0047 .0049 .0018 .0140 .0029 .0045
122 .0000 .0000 .0000 .0000 .0000 .0000 .0046 .0050 .0022 .0029 .0054 .0064 .0033 .0262 .0098 .0154 .0053 .0055 .0049 .0057 .0073 .0049 .0052 .0046 .0043 .0070 .0054
123 .0000 .0000 .0000 .0000 .0000 .0066 .0023 .0043 .0039 .0032 .0034 .0074 .0050 .0254 .0072 .0183 .0074 .0047 .0303 .0068 .0055 .0039 .0051 .0283 .0058 .0106 .0055
124 .0000 .0000 .0000 .0000 .0000 .0070 .0026 .0016 .0020 .0018 .0058 .0063 .0042 .0290 .0071 .0075 .0034 .0058 .0048 .0068 .0073 .0036 .0069 .0078 .0031 .0054 .0035
126 .0000 .0000 .0000 .0000 .0000 .0017 .0084 .0069 .0092 .0075 .0044 .0117 .0145 .0266 .0159 .0286 .0139 .0064 .0149 .0138 .0128 .0085 .0120 .0154 .0203 .0285 .0095
139 .0000 .0000 .0000 .0000 .0000 .0010 .0010 .0010 .0008 .0013 .0011 .0015 .0015 .0007 .0013 .0021 .0018 .0012 .0008 .0012 .0016 .0015 .0015 .0016 .0011 .0006 .0010
142 .0000 .0000 .0000 .0000 .0000 .0000 .0155 .0191 .0139 .0040 .0000 .0121 .0297 .0098 .0236 .0325 .0188 .0083 .0085 .0088 .0142 .0140 .0206 .0258 .0125 .0166 .0200
146 .0000 .0000 .0000 .0000 .0000 .0000 .0030 .0019 .0030 .0029 .0020 .0050 .0038 .0203 .0048 .0215 .0149 .0042 .0044 .0047 .0032 .0037 .0042 .0201 .0030 .0039 .0032
150 .0000 .0000 .0000 .0000 .0000 .0003 .0013 .0005 .0010 .0014 .0010 .0012 .0010 .0005 .0003 .0015 .0017 .0009 .0011 .0015 .0009 .0007 .0014 .0015 .0000 .0015 .0033
151 .0000 .0000 .0000 .0000 .0000 .0010 .0200 .0043 .0034 .0036 .0039 .0150 .0085 .0006 .0031 .0080 .0069 .0149 .0080 .0178 .0062 .0254 .0257 .0035 .0139 .0079 .0031
157 .0000 .0000 .0000 .0000 .0000 .0365 .0013 .0029 .0020 .0020 .0024 .0039 .0033 .0592 .0029 .0084 .0029 .0034 .0026 .0020 .0031 .0023 .0042 .0056 .0038 .0066 .0082
159 .0000 .0000 .0000 .0000 .0000 .0000 .0007 .0004 .0006 .0006 .0030 .0006 .0007 .0004 .0017 .0016 .0014 .0007 .0012 .0005 .0008 .0007 .0009 .0007 .0021 .0008 .0008
160 .0000 .0000 .0000 .0000 .0000 .0021 .0012 .0010 .0010 .0016 .0021 .0015 .0018 .0009 .0021 .0056 .0030 .0023 .0013 .0015 .0020 .0017 .0014 .0021 .0018 .0015 .0045
163 .0000 .0000 .0000 .0000 .0000 .0000 .0009 .0019 .0009 .0006 .0020 .0015 .0018 .0008 .0026 .0060 .0018 .0018 .0015 .0008 .0009 .0011 .0014 .0014 .0013 .0029 .0015
164 .0000 .0000 .0000 .0000 .0000 .0011 .0020 .0018 .0014 .0013 .0010 .0012 .0011 .0008 .0016 .0019 .0021 .0016 .0020 .0015 .0015 .0009 .0029 .0016 .0013 .0024 .0017
166 .0000 .0000 .0000 .0000 .0000 .0000 .0025 .0008 .0000 .0010 .0000 .0016 .0021 .0000 .0017 .0010 .0012 .0011 .0013 .0026 .0017 .0010 .0009 .0000 .0000 .0003 .0012
170 .0000 .0000 .0000 .0000 .0000 .0000 .0029 .0020 .0027 .0021 .0114 .0034 .0082 .0056 .1151 .0314 .0047 .0041 .0028 .0032 .0056 .0027 .0034 .0059 .0901 .0175 .0048
171 .0000 .0000 .0000 .0000 .0000 .0024 .0012 .0018 .0015 .0025 .0031 .0037 .0046 .0022 .0075 .0092 .0068 .0042 .0048 .0013 .0027 .0030 .0046 .0031 .0035 .0031 .0008
175 .0000 .0000 .0000 .0000 .0000 .0088 .0056 .0099 .0070 .0075 .0064 .0085 .0071 .0057 .0096 .0166 .0141 .0079 .0072 .0122 .0094 .0073 .0096 .0073 .0089 .0096 .0111
182 .0000 .0000 .0000 .0000 .0000 .0000 .0051 .0032 .0049 .0039 - .0047 .0085 .0041 .0331 .0294 .0042 .0067 .0060 .0075 OOM .0039 .0050 .0048 .0085 .0056 .0036
183 .0000 .0000 .0000 .0000 .0000 .0000 .0097 .0151 .0059 .0048 .0055 .0374 .0057 .0050 .0100 .0201 .0085 .0394 .0098 .0390 .0101 .0205 .0082 .0220 .0120 .0097 .0068
185 .0000 .0000 .0000 .0000 .0000 .0000 .0032 .0016 .0027 .0028 .0030 .0036 .0054 .0013 .0050 .0052 .0028 .0034 .0024 .0025 .0031 .0069 .0034 .0035 .0033 .0041 .0035
186 .0000 .0000 .0000 .0000 .0000 .0666 .0205 .0169 .0112 .0149 .0192 .0716 .0898 .0152 .0286 .0222 .0410 .0148 .0445 .0159 .0561 .0672 .0777 .0435 .0507 .0340 .1730
187 .0000 .0000 .0000 .0000 .0000 .0053 .0023 .0015 .0018 .0012 .0031 .0033 .0039 .0027 .0143 .0113 .0042 .0031 .0029 .0021 .0016 .0032 .0035 .0034 .0022 .0036 .0023
193 .0000 .0000 .0000 .0000 .0000 .0025 .0006 .0009 .0003 .0005 .0008 .0009 .0010 .0033 .0010 .0011 .0007 .0005 .0008 .0005 .0010 .0007 .0011 .0012 .0006 .0013 .0016
194 .0000 .0000 .0000 .0000 .0000 .0000 .0000 .0000 .0000 .0000 .0024 .0082 .0080 .0097 .0149 .0565 .0098 .0050 .0076 .0000 .0624 .0068 .0098 .0073 .0499 .0227 .0000
195 .0000 .0000 .0000 .0000 .0000 .0056 .0025 .0032 .0042 .0035 .0064 .0069 .0075 .0041 .0079 .0154 .0064 .0054 .0078 .0069 .0059 .0183 .0060 .0051 .0474 .0037 .0059
198 .0000 .0000 .0000 .0000 .0000 .0017 .0023 .0014 .0016 .0024 .0025 .0055 .0041 .0019 .0175 .0039 .0026 .0052 .0046 .0017 .0019 .0052 .0044 .0051 .0043 .0026 .0022
200 .0000 .0000 .0000 .0000 .0000 .0000 .0045 .0061 .0036 .0051 .0015 .0063 .0058 .0054 .0070 .0069 .0080 .0041 .0033 .0043 .0047 .0090 .0056 .0112 .0048 .0084 .0062
201 .0000 .0000 .0000 .0000 .0000 .0039 .0024 .0008 .0021 .0021 .0040 .0039 .0025 .0006 .0052 .0038 .0044 .0024 .0030 .0048 .0026 .0036 .0038 .0043 .0109 .0042 .0046
33
209 .0000 .0000 .0000 .0000 .0000 .0000 .0076 .0050 .0144 .0051 - .0139 .0180 .0096 .0213 .0839 .0231 .0109 .0132 .0295 .0118 .0368 .0149 .0168 .0168 .0154 .0485
210 .0000 .0000 .0000 .0000 .0000 .0007 .0014 .0005 .0004 .0008 .0007 .0010 .0005 .0023 .0092 .0016 .0020 .0013 .0007 .0008 .0008 .0010 .0010 .0013 .0008 .0009 .0008
221 .0000 .0000 .0000 .0000 .0000 .0084 .0005 .0004 .0004 .0000 .0034 .0010 .0020 .0011 .0031 .0026 .0011 .0016 .0015 .0008 .0012 .0009 .0012 .0040 .0009 .0080 .0006
222 .0000 .0000 .0000 .0000 .0000 .0416 .0013 .0009 .0007 .0011 .0039 .0039 .0044 .0004 .0225 .0054 .0030 .0109 .0042 .0019 .0088 .0022 .0821 .0056 .0035 .0037 .0013
224 .0000 .0000 .0000 .0000 .0000 .0007 .0006 .0009 .0009 .0008 .0017 .0008 .0012 .0016 .0019 .0034 .0008 .0008 .0009 .0008 .0012 .0011 .0043 .0010 .0023 .0010 .0012
225 .0000 .0000 .0000 .0000 .0000 .0025 .0000 .0000 .0000 .0000 .0009 .0041 .0057 .0008 .0080 .0064 .0042 .0049 .0025 .0019 .0014 .0051 .0063 .0270 .0043 .0142 .0032
230 .0000 .0000 .0000 .0000 .0000 .0033 .0021 .0020 .0015 .0017 .0020 .0025 .0015 .0019 .0025 .0029 .0017 .0015 .0013 .0029 .0028 .0037 .0043 .0024 .0049 .0059 .0019
232 .0000 .0000 .0000 .0000 .0000 .0000 .0038 .0031 .0019 .0034 .0025 .0077 .0054 .0000 .0072 .0024 .0056 .0046 .0076 .0033 .0041 .0048 .0061 .0060 .0083 .0043 .0042
235 .0000 .0000 .0000 .0000 .0000 .0000 .0031 .0065 .0063 .0040 .0000 .0101 .0080 .0000 .0109 .0088 .0035 .0052 .0069 .0063 .0000 .0075 .0073 .0000 .0015 .0098 .0042
236 .0000 .0000 .0000 .0000 .0000 .0095 .0044 .0062 .0031 .0052 .0038 .0044 .0079 .0000 .0128 .0089 .0045 .0062 .0070 .0029 .0039 .0055 .0049 .0008 .0063 .0067 .0045
237 .0000 .0000 .0000 .0000 .0000 .0027 .0056 .0081 .0026 .0000 .0023 .0057 .0117 .0030 .0191 .0064 .0054 .0080 .0048 .0078 .0040 .0071 .0063 .0073 .0046 .0091 .0043
239 .0000 .0000 .0000 .0000 .0000 .0253 .0053 .0049 .0038 .0000 .0070 .0068 .0037 .0026 .0072 .0117 .0139 .0043 .0097 .0061 .0126 .0062 .0083 .0069 .0062 .0068 .0042
240 .0000 .0000 .0000 .0000 .0000 .0000 .0014 .0009 .0018 .0009 .0017 .0024 .0023 .0009 .0042 .0043 .0024 .0014 .0012 .0012 .0010 .0013 .0029 .0022 .0020 .0445 .0017
246 .0000 .0000 .0000 .0000 .0000 .0063 .0102 .0037 .0045 .0094 .0064 .0158 .0198 .0307 .0160 .0474 .0183 .0078 .0073 .0104 .0129 .0081 .0057 .0116 .0092 .0184 .0101
249 .0000 .0000 .0000 .0000 .0000 .0000 .0015 .0023 .0021 .0011 .0013 .0022 .0036 .0032 .0026 .0048 .0029 .0025 .0028 .0027 .0029 .0024 .0035 .0025 .1526 .0041 .0034
250 .0000 .0000 .0000 .0000 .0000 .0030 .0014 .0017 .0026 .0017 .0014 .0026 .0030 .0010 .0107 .0133 .0044 .0024 .1380 .0027 .0040 .0029 .0038 .0102 .0036 .0018 .0011
251 .0000 .0000 .0000 .0000 .0000 .0191 .0077 .0073 .0036 .0050 .0066 .0120 .0142 .0247 .0094 .0242 .0306 .0108 .0171 .0145 .0102 .0118 .0140 .0120 .0116 .0172 .0035
255 .0000 .0000 .0000 .0000 .0000 .0042 .0023 .0048 .0007 .0020 .0000 .0073 .0066 .0010 .0059 .0050 .0012 .0058 .0073 .0054 .0061 .0039 .0032 .0036 .0046 .0039 .0064
263 .0000 .0000 .0000 .0000 .0000 .0037 .0023 .0030 .0022 .0027 .0015 .0045 .0033 .0020 .0047 .0081 .0028 .0041 .0051 .0039 .0040 .0030 .0073 .0020 .0043 .0048 .0064
265 .0000 .0000 .0000 .0000 .0000 .0042 .0057 .0052 .0069 .0037 .0037 .0062 .0085 .0474 .0241 .0369 .0046 .0107 .0113 .0146 .0122 .0109 .0196 .0196 .0089 .0103 .0082
266 .0000 .0000 .0000 .0000 .0000 .0000 .0135 .0146 .0106 .0106 .0066 .0180 .0214 .0095 .0220 .0310 .0213 .0085 .0075 .0168 .0130 .0146 .0197 .0142 .0178 .0265 .0115
273 .0000 .0000 .0000 .0000 .0000 .0000 .0019 .0024 .0019 .0000 .0026 .0061 .0027 .0021 .0080 .0080 .0041 .0038 .0048 .0055 .0060 .0023 .0935 .0066 .0030 .0080 .0030
274 .0000 .0000 .0000 .0000 .0000 .0000 .0024 .0013 .0008 .0016 .0013 .0014 .0011 .0002 .0462 .0014 .0016 .0009 .0024 .0010 .0015 .0019 .0037 .0017 .0023 .0021 .0025
275 .0000 .0000 .0000 .0000 .0000 .0034 .0022 .0016 .0034 .0038 .0027 .0030 .0029 .0021 .0068 .0075 .0040 .0030 .0126 .0044 .0026 .0027 .0047 .0036 .0017 .0040 .0045
283 .0000 .0000 .0000 .0000 .0000 .0000 .0020 .0010 .0015 .0022 .0014 .0018 .0034 .0006 .0043 .0089 .0016 .0022 .0011 .0009 .0025 .0047 .0017 .0081 .0025 .0340 .0015
289 .0000 .0000 .0000 .0000 .0000 .0000 .0030 .0008 .0017 .0018 .0019 .0022 .0024 .0012 .0066 .0040 .0057 .0026 .0026 .0070 .0048 .0041 .0028 .0023 .0029 .0041 .0027
296 .0000 .0000 .0000 .0000 .0000 .0000 .0104 .0079 .0047 .0000 .0043 .0098 .0087 .0127 .0104 .0103 .0108 .0052 .0063 .0106 .0059 .0099 .0079 .0077 .0082 .0115 .0130
Table 17: Additional results of multi-class classification datasets. For each dataset-method pair,
we list the standard deviation of accuracy on 15 random seeds. The column names continue the
denotation in Table 16.
5 .0000 .0000 .0000 .0000 .0000 .0026 .0113 .0215 .0110 .0077 - .0246 .0147 .0128 .0188 .0550 .0219 .0217 .0231 .0416 .0344 .0190 .0434 .0337 .0350 .0243 .0107
10 .0000 .0000 .0000 .0000 .0000 .0000 .0044 .0028 .0035 .0039 .0035 .0038 .0029 .0007 .0046 .0028 .0043 .0018 .0039 .0039 .0028 .0026 .0027 .0096 .0031 .0051 .0036
12 .0000 .0000 .0000 .0000 .0000 .0000 .0014 .0012 .0012 .0009 .0027 .0021 .0016 .0010 .0021 .0040 .0018 .0014 .0013 .0014 .0013 .0016 .0018 .0009 .0350 .0038 .0014
36 .0000 .0000 .0000 .0000 .0000 .0009 .0015 .0015 .0019 .0020 .0051 .0044 .0018 .0008 .0023 .0012 .0046 .0030 .0025 .0031 .0020 .0027 .0018 .0020 .0027 .0024 .0014
39 .0000 .0000 .0000 .0000 .0000 .0000 .0002 .0002 .0008 .0010 .0045 .0005 .0006 .0006 .0063 .0035 .0023 .0007 .0019 .0003 .0009 .0005 .0003 .0008 .0007 OOM OOM
55 .0000 .0000 .0000 .0000 .0000 .0050 .0032 .0014 .0037 .0024 .0036 .0029 .0033 .0265 .0062 .0083 .0036 .0035 .0049 .0034 .0049 .0060 .0046 .0033 .0051 .0027 .0045
60 .0000 .0000 .0000 .0000 .0000 .0051 .0051 .0084 .0035 .0026 .0022 .0051 .0045 .0016 .0093 .0074 .0061 .0038 .0048 .0035 .0048 .0045 .0055 .0081 .0062 .0054 .0048
61 .0000 .0000 .0000 .0000 .0000 .0093 .0041 .0029 .0043 .0055 .0076 .0118 .0122 .0131 .0126 .0149 .0110 .0119 .0112 .0184 .0099 .0082 .0106 .0063 .0532 .0143 .0053
68 .0000 .0000 .0000 .0000 .0000 .0078 .0031 .0033 .0035 .1294 - .0051 .0035 .0738 .0466 .0084 .0093 .0031 .0054 .0045 .0025 .0034 .0056 .0035 .0064 .0851 .0023
72 .0000 .0000 .0000 .0000 .0000 .0017 .0019 .0010 .0018 .0014 .0041 .0018 .0014 .0026 .0066 .0036 .0032 .0026 .0016 .0025 .0017 .0017 .0011 .0045 .0025 .0032 .0014
88 .0000 .0000 .0000 .0000 .0000 .0096 .0045 .0053 .0093 .0047 .0030 .1149 .0128 .0617 .0232 .0599 .0120 .0093 .1788 .0984 .0122 .0046 .0102 .1540 .0124 .0247 .0066
102 .0000 .0000 .0000 .0000 .0000 .0000 .0007 .0007 .0005 .0006 .0016 .0007 .0007 .0007 .0011 .0036 .0012 .0010 .0009 .0019 .0020 .0019 .0026 .0015 .0018 .0011 .0412
103 .0000 .0000 .0000 .0000 .0000 .0093 .0002 .0003 .0006 .0002 .0017 .0005 .0007 .0002 .0298 .0010 .0010 .0005 .0006 .0006 .0143 .0006 .0006 .0013 .0009 .0048 .0005
112 .0000 .0000 .0000 .0000 .0000 .0000 .0096 .0192 .0170 .0151 - .0223 .0208 .0272 .0262 .0487 .0353 .1899 .0281 .0401 .0264 .0158 .0232 .0141 .1485 .0299 .0233
120 .0000 .0000 .0000 .0000 .0000 .0326 .0047 .0062 .0055 .0057 .0051 .0079 .0094 .0145 .0230 .0131 .0049 .0042 .0054 .0046 .0083 .0049 .0090 .0075 .0058 .0111 .0045
121 .0000 .0000 .0000 .0000 .0000 .0313 .0010 .0022 .0013 .0000 - .0028 .0021 OOM .0150 .0072 .0036 .0026 .0015 .0013 .0025 .0015 .0015 .0062 .0022 .0044 .0012
127 .0000 .0000 .0000 .0000 .0000 .0039 .0012 .0020 .0014 .0016 .0010 .0012 .0021 .0011 .0019 .0047 .0022 .0016 .0014 .0026 .0024 .0020 .0024 .0033 .0068 .0020 .0036
128 .0000 .0000 .0000 .0000 .0000 .0174 .0017 .0036 .0020 .0012 .0016 .0028 .0041 .0000 .0034 .0011 .0042 .0037 .0025 .0032 .0026 .0051 .0054 .0013 .0000 .0045 .0035
129 .0000 .0000 .0000 .0000 .0000 .0000 .0015 .0036 .0026 .0030 .0015 .0054 .0480 .0156 .0026 .0338 .0026 .0036 .0041 .0591 .0060 .0062 .0048 .0026 .0096 .0115 .0051
134 .0000 .0000 .0000 .0000 .0000 .0081 .0030 .0027 .0037 .0036 .0082 .0146 .0205 .0036 .0136 .0172 .0320 .0092 .0279 .0094 .0371 .0125 .0246 .0138 .0157 .0085 .0064
136 .0000 .0000 .0000 .0000 .0000 .0014 .0090 .0176 .0185 .0059 .0137 .0172 .0135 .0200 .0238 .0218 .0169 .0120 .0097 .0189 .0170 .0205 .0674 .0163 .0304 .0177 .0163
137 .0000 .0000 .0000 .0000 .0000 .0000 .0109 .0176 .0114 .0098 .0130 .0175 .0229 .0116 .0226 .0421 .0240 .0144 .0173 .0194 .0272 .0119 .0172 .0103 .0288 .0363 .0175
143 .0000 .0000 .0000 .0000 .0000 .0000 .0074 .0030 .0033 .0020 .0013 .0044 .0076 .0068 .0111 .0138 .0060 .0043 .0095 .0054 .0111 .0097 .0082 .0088 .0100 .0040 .0059
145 .0000 .0000 .0000 .0000 .0000 .0114 .0031 .0031 .0045 .0028 .0035 .0116 .0102 .0684 .0131 .0465 .0075 .0071 .0075 .0190 .0084 .0093 .0069 .0029 .0422 .1308 .0086
147 .0000 .0000 .0000 .0000 .0000 .0094 .0086 .0141 .0090 .0038 .0062 .0202 .0257 .0130 .0208 .0471 .0173 .0152 .0208 .0129 .0134 .0126 .0249 .0169 .0624 .0264 .0139
154 .0000 .0000 .0000 .0000 .0000 .0245 .0080 .0060 .0116 .0028 .0087 .0136 .0224 .0111 .0218 .0404 .0182 .0130 .0284 .0636 .0208 .0237 .0248 .0136 .0515 .0196 .0259
167 .0000 .0000 .0000 .0000 .0000 .0000 .0027 .0021 .0047 .0030 - .0087 .0051 .0917 .0071 .0534 .0091 .0065 .0074 .1118 .0046 .0060 .0039 .0058 .0748 OOM .0058
168 .0000 .0000 .0000 .0000 .0000 .0000 .0079 .0111 .0073 .0063 .0000 .0152 .0174 .0085 .0081 .0131 .0132 .0230 .0115 .0075 .0169 .0145 .0155 .0147 .0187 .0186 .0138
169 .0000 .0000 .0000 .0000 .0000 .0032 .0013 .0018 .0020 .0018 .0026 .0022 .0015 .0013 .0066 .0044 .0014 .0019 .0022 .0022 .0022 .0022 .0020 .0028 .0027 .0415 .0014
173 .0000 .0000 .0000 .0000 .0000 .0126 .0053 .0035 .0056 .0040 .0058 .0071 .0052 .0560 .0316 .0353 .0061 .0056 .0057 .0046 .0063 .0056 .0101 .0098 .0067 .0325 .0039
174 .0000 .0000 .0000 .0000 .0000 .0000 .0064 .0034 .0054 .0059 .0093 .0054 .0142 .0071 .0071 .0384 .0150 .0080 .0068 .0135 .0105 .0088 .0231 .0133 .0239 .0110 .0051
177 .0000 .0000 .0000 .0000 .0000 .0000 .0055 .0045 .0042 .0047 .0072 .0076 .0067 .0350 .0200 .0286 .0084 .0077 .0062 .0080 .0110 .0096 .0091 .0459 .0068 .0135 .0064
180 .0000 .0000 .0000 .0000 .0000 .0007 .0004 .0005 .0002 .0005 - .0008 .0008 .0008 .0950 .0073 .0016 .0014 .0007 .0011 .0010 .0009 .0010 .0013 .0012 .0461 .0004
196 .0000 .0000 .0000 .0000 .0000 .0123 .0000 .0001 .0000 .0000 .0390 .0019 .0544 .0002 .0002 .0770 .0012 .0000 .0015 .0011 .0020 .0021 .0016 .0881 .0001 .0580 .0002
197 .0000 .0000 .0000 .0000 .0000 .0000 .0034 .0046 .0032 .0026 - .0064 .0069 .0052 .0076 .0271 .0138 .0071 .0081 .0071 .0081 .0043 .0088 .0100 .0711 .0055 .0051
199 .0000 .0000 .0000 .0000 .0000 .0003 .0018 .0010 .0012 .0000 .0037 .0146 .0042 .0009 .0124 .0142 .0101 .0066 .0200 .0153 .0161 .0051 .0216 .0058 .0048 .0124 .0095
203 .0000 .0000 .0000 .0000 .0000 .0000 .3661 .0051 .0035 .0000 - .0084 .0051 .0019 .0264 .0100 .0097 .0040 .0115 .0036 .0074 .0049 .0043 .0050 .0072 .0081 .0045
204 .0000 .0000 .0000 .0000 .0000 .0001 .0021 .0044 .0030 .0000 - .0052 .0059 .0018 .0253 .0113 .0183 .0042 .0095 .0066 .0054 .0040 .0153 .0057 .0068 .0090 .0035
206 .0000 .0000 .0000 .0000 .0000 .0000 .0042 .0007 .0053 .0032 .0094 .0063 .0054 .0033 .0087 .0131 .0074 .0048 .0068 .0048 .0073 .0053 .0067 .0059 .0119 .0074 .0034
207 .0000 .0000 .0000 .0000 .0000 .0000 .0039 .0008 .0024 .0050 .0048 .1611 .0035 .0065 .0038 .0135 .0056 .0044 .0032 .0031 .0071 .0038 .0037 .0067 .0049 .0029 .0030
208 .0000 .0000 .0000 .0000 .0000 .0013 .0018 .0013 .0030 .0000 - .0017 .0011 .0014 .0102 .0063 .0044 .0030 .0030 .0033 .0023 .0016 .0029 .0039 .0030 .0097 .0013
211 .0000 .0000 .0000 .0000 .0000 .0000 .0099 .0102 .0065 .0067 .0095 .0462 .0478 .0286 .0335 .0361 .0176 .0266 .0376 .0246 .0207 .0243 .0323 .0234 .0150 .0105 .0137
213 .0000 .0000 .0000 .0000 .0000 .0000 .0022 .0026 .0023 .0015 - .0062 .0040 .0318 .0065 .0184 .0024 .0024 .0053 .0035 .0031 .0046 .0036 .0031 .0042 .0037 .0036
214 .0000 .0000 .0000 .0000 .0000 .0000 .0061 .0048 .0101 .0061 .0183 .0079 .0087 .0070 .0194 .0556 .0095 .0111 .0119 .0117 .0114 .0095 .0097 .0108 .0105 .0118 .0172
215 .0000 .0000 .0000 .0000 .0000 .0000 .0039 .0021 .0026 .0028 .0109 .0046 .0032 .0081 .0055 .0504 .0066 .0066 .0036 .0046 .0076 .0069 .0049 .0071 .0071 .0043 .0038
216 .0000 .0000 .0000 .0000 .0000 .0082 .0030 .0046 .0083 .0032 .0080 .0117 .0098 .0099 .0580 .0205 .0104 .0112 .0093 .0067 .0093 .0114 .0073 .0108 .0089 .0116 .0103
217 .0000 .0000 .0000 .0000 .0000 .0000 .0033 .0040 .0016 .0022 - .0051 .0031 .0037 .0046 .0264 .0063 .0041 .0066 .0082 .0035 .0022 .0062 OOM .0080 OOM .0045
218 .0000 .0000 .0000 .0000 .0000 .0066 .0114 .0084 .0084 .0071 .0120 .0128 .0112 .0365 .0115 .0462 .0099 .0107 .0139 .0084 .0136 .0060 .0087 .0089 .0216 .0101 .0062
219 .0000 .0000 .0000 .0000 .0000 .0000 .0048 .0045 .0050 .0034 .0047 .0047 .0016 .0041 .0128 .0425 .0033 .0019 .0050 .0080 .0073 .0061 .0041 .0047 .0078 .0029 .0027
220 .0000 .0000 .0000 .0000 .0000 .0006 .0020 .0021 .0017 .0021 .0048 .0032 .0019 .0011 .0028 .0053 .0040 .0031 .0023 .0020 .0028 .0022 .0036 .0158 .0029 .0608 .0025
226 .0000 .0000 .0000 .0000 .0000 .0027 .0014 .0019 .0012 .0014 .0015 .0010 .0014 .0014 .0029 .0033 .0020 .0013 .0021 .0016 .0029 .0019 .0014 .0034 .0015 .0028 .0014
227 .0000 .0000 .0000 .0000 .0000 .0000 .0128 .0078 .0130 .0119 - .0102 .0120 .0129 .0259 .1726 .0269 .0217 .0142 .0188 .0120 .0186 .0147 .0227 .0244 OOM .0176
228 .0000 .0000 .0000 .0000 .0000 .0068 .0112 .0130 .0149 .0083 - .0139 .0160 .0347 .0419 .2012 .0176 .0257 .0353 .0247 .0202 .0145 .0170 .0176 .0375 .0179 .0141
34
229 .0000 .0000 .0000 .0000 .0000 .0000 .0104 .0085 .0133 .0000 - .0126 .0099 .0250 .0107 .1824 .0098 .0112 .0147 .0085 .0121 .0120 .0098 .0210 .0262 OOM .0124
231 .0000 .0000 .0000 .0000 .0000 .0000 .0022 .0011 .0018 .0017 .0034 .0016 .0015 .0059 .0032 .0089 .0016 .0022 .0021 .0032 .0020 .0026 .0019 .0023 .0020 .0025 .0020
234 .0000 .0000 .0000 .0000 .0000 .0020 .0013 .0028 .0014 .0015 .0018 .0033 .0029 .0256 .0845 .0056 .0016 .0015 .0023 .0032 .0025 .0025 .0038 .0032 .0035 .0030 .0023
238 .0000 .0000 .0000 .0000 .0000 .0013 .1887 .0007 .0010 .1431 .0031 .0008 .0010 .0004 .0020 .0034 .0009 .0009 .0016 .0012 .0007 .0010 .0020 .0013 .0016 .0366 .0005
243 .0000 .0000 .0000 .0000 .0000 .0128 .0033 .0054 .0047 .0049 .0050 .0060 .0094 .0065 .0068 .0186 .0088 .0107 .0051 .0055 .0068 .0034 .0039 .0056 .0105 .0072 .0038
253 .0000 .0000 .0000 .0000 .0000 .0037 .0030 .0025 .0028 .0036 .0027 .0090 .0075 .0296 .0083 .0070 .0092 .0071 .0096 .0100 .0079 .0062 .0063 .0060 .0073 .0051 .0036
254 .0000 .0000 .0000 .0000 .0000 .0270 .0034 .0064 .0037 .0028 .0053 .0058 .0041 .0281 .0348 .0161 .0140 .0115 .0087 .0088 .0068 .0062 .0215 .0050 .0107 .0601 .0053
256 .0000 .0000 .0000 .0000 .0000 .0000 .0071 .0052 .0080 .0079 - .0087 .0111 .0168 .0112 .0370 .0080 .0076 .0089 .0058 .0063 .0111 .0077 .0069 .0112 .0086 .0082
260 .0000 .0000 .0000 .0000 .0000 .0199 .0001 .0000 .0000 .1491 - .0003 .0003 .0000 .0185 .0007 .0004 .0007 .0002 .0002 .0002 .0002 .0002 .0011 .0003 .0102 .0001
264 .0000 .0000 .0000 .0000 .0000 .0091 .0045 .0028 .0036 .0026 .0042 .0103 .0115 .1743 .0093 .0329 .0094 .0104 .0100 .0044 .0067 .0111 .0053 .0039 .0054 .0584 .0111
268 .0000 .0000 .0000 .0000 .0000 .0068 .0125 .0050 .0064 .0058 .0101 .0099 .0134 .0463 .0517 .0322 .0071 .0101 .0083 .0131 .0087 .0457 .0082 .0433 .0102 .0615 .0084
276 .0000 .0000 .0000 .0000 .0000 .0010 .0018 .0019 .0025 .2068 - .0019 .0011 .0504 .0059 .0075 .0015 .0008 .0024 .0016 .0014 .0013 .0027 .0018 .0021 .0430 .0008
277 .0000 .0000 .0000 .0000 .0000 .0143 .0007 .0003 .0006 .0004 .0020 .0018 .0022 .0250 .0053 .0039 .0026 .0015 .0015 .0016 .0012 .0016 .0018 .0061 .0027 .0029 .0019
278 .0000 .0000 .0000 .0000 .0000 .0121 .0014 .0006 .0014 .0000 .0015 .0018 .0026 .0127 .0052 .0169 .0041 .0024 .0026 .0016 .0019 .0095 .0028 .0029 .0141 .0044 .0033
279 .0000 .0000 .0000 .0000 .0000 .0070 .0027 .0032 .0029 .0042 .0050 .0064 .0091 .0060 .0093 .0305 .0087 .0044 .0043 .0084 .0380 .0116 .0046 .0089 .0081 .0064 .0070
282 .0000 .0000 .0000 .0000 .0000 .0194 .0034 .0034 .0050 .0020 .0025 .0100 .0099 .0042 .0104 .0053 .0082 .0174 .0103 .0030 .0130 .0041 .0089 .0121 .0048 .0132 .0043
284 .0000 .0000 .0000 .0000 .0000 .0387 .0108 .0222 .0157 .0141 .0060 .0313 .0109 .0400 .0240 .0505 .0162 .0467 .0316 .0273 .0237 .0185 .0204 .0131 .0377 .0162 .0160
285 .0000 .0000 .0000 .0000 .0000 .0013 .0011 .0021 .0006 .0016 - .0021 .0027 .0018 .0028 .0070 .0038 .0024 .0044 .0046 .0038 .0032 .0020 .0033 .0025 .0048 OOM
287 .0000 .0000 .0000 .0000 .0000 .0150 .0014 .0012 .0010 .0000 - .0019 .0025 OOM .0118 .0043 .0063 .0024 .0029 .0019 .0030 .0033 .0023 .0028 .0020 .0094 .0009
288 .0000 .0000 .0000 .0000 .0000 .0041 .0010 .0009 .0011 .0000 .0027 .0055 .0038 .0033 .0086 .0170 .0091 .0060 .0062 .0060 .0056 .0033 .0044 .0076 .0060 .0040 .0021
290 .0000 .0000 .0000 .0000 .0000 .0035 .0036 .0072 .0048 .0037 .0057 .0078 .0038 .0063 .0051 .0127 .0061 .0057 .0037 .0049 .0082 .0048 .0064 .0057 .0776 .0331 .0050
291 .0000 .0000 .0000 .0000 .0000 .0131 .0109 .0166 .0114 .0063 .0056 .0082 .0076 .0086 .0076 .0122 .0136 .0082 .0084 .0111 .0103 .0129 .0113 .0127 .0083 .0093 .0119
292 .0000 .0000 .0000 .0000 .0000 .0057 .0030 .0044 .0042 .0034 .0031 .0060 .0042 .0017 .0065 .0129 .0062 .0043 .0046 .0069 .0045 .0056 .0048 .0069 .0067 .0049 .0042
294 .0000 .0000 .0000 .0000 .0000 .0000 .0048 .0092 .0069 .0046 .0065 .0128 .0099 .0048 .0090 .0218 .0151 .0154 .0119 .0252 .0184 .0090 .0211 .0129 .1170 .0534 .0085
298 .0000 .0000 .0000 .0000 .0000 .0160 .0100 .0045 .0084 .0091 .0076 .0220 .0200 .0140 .0444 .0195 .0167 .0280 .0429 .0107 .0234 .0097 .0140 .0076 .0082 .0192 .0173
299 .0000 .0000 .0000 .0000 .0000 .0000 .0061 .0052 .0067 .0049 .0073 .0207 .0178 .0163 .0092 .0266 .0097 .0264 .0118 .0128 .0090 .0073 .0155 .0089 .0123 .0690 .0129
300 .0000 .0000 .0000 .0000 .0000 .0125 .0072 .0063 .0085 .0000 .0165 .0154 .0075 .0061 .0234 .0511 .0133 .0132 .0150 .0108 .0126 .0108 .0155 .0151 .0110 .0772 .0122
Table 18: Additional results of regression datasets. or each dataset-method pair, we list the standard
deviation of RMSE on 15 random seeds. ID continues the ID in Table 4. “DU” means Dummy as
the naive baseline, “LR”, “XGB”, “CB”, “RF”, “LG”, “RN”, “ND”, “ST”, “TN”, “TG”, “DAN”,
“FT”, “AI”, “DCN”, “TT”, “PT”, “GN”, “TR”, and “DR” denote “Linear Regression”, “XGBoost”,
“CatBoost”, “Random Forest”, “LightGBM”, “ResNet”, “NODE”, “SwitchTab”, “TabNet”, “TAN-
GOS”, “DANets”, “FT-T”, “AutoInt”, “DCNv2”, “TabTransformer”, “PTaRL”, “GrowNet”, “TabR”,
and “DNNR”, respectively. In each cell, if the experiment is failed to perform due to out of memory,
we denote it by “OOM”. The scientific notation beside the dataset name indicates the scale of the
results, e.g., ×10 means the final value of all results should be multiple by 10. And if the standard
deviation of a method is too large compared to others, i.e., still exceeds 1 after scaling, we denote it
by “INF”.
ID DU LR KNN SVM XGB CB RF LG MLP RN ND ST TN TG DAN FT AI DCN SNN TT PT GN TR DR
3
1 (×10 ) .0000 .0000 .0000 .0384 .0262 .0040 .0058 .0061 .0418 .0723 .0101 .0280 .0636 .0250 INF .0589 .0567 .0436 .0345 .0152 .0673 .0225 .0353 .0094
2 (×10−1 ) .0000 .0000 .0000 .0036 .0152 .0023 .0040 .0000 .0172 .0134 .0042 .6492 .2778 .0200 INF .0114 .0341 .0039 .0275 .2528 .0107 .0955 .0105 .0690
3 (×10−1 ) .0000 .0000 .0000 .1031 .0283 .0050 .0729 .0791 .2097 .1491 .0174 .0014 .5076 .0579 .0019 .1655 .2541 .1711 .1084 .0035 INF .1247 .0005 .0000
4 (×10−1 ) .0000 .0000 .0000 .0000 .0067 .0007 .0403 .0065 .2980 .3726 .0327 .0048 .3084 .0546 .0090 .5587 .1558 .2220 .4122 .0078 .4723 .0548 .0004 .0310
6 (×100 ) .0000 .0000 .0000 .1002 .0085 .0170 .0045 .0000 .0119 .0069 .0044 .1784 .0427 .0085 .1565 .0101 .0128 .0104 .0112 .0375 .0127 .0476 .0236 .0150
7 (×10−4 ) .0000 .0000 .0000 .0200 .0099 .0042 .0052 .0035 .0221 .0308 .0020 .0697 .0691 .0131 .2149 .0186 .0135 .9263 .0647 .0230 .0112 INF .0099 .0167
9 (×102 ) .0000 .0000 .0000 .0185 .0623 .2960 .0495 .0313 .0304 .0802 .0513 INF .2403 .1594 INF .1939 .1585 .0853 .1227 .3762 .0379 .4283 .1635 .0546
13 (×100 ) .0000 .0000 .0000 .0014 .0184 .0113 .0100 .0109 .0311 .0297 .0055 .1130 .0826 .0423 .4607 .1139 .0473 .0248 .0655 INF .0413 .0357 .0333 .0483
14 (×101 ) .0000 .0000 .0000 .1887 .0309 .0151 .0187 .0142 .1078 .0872 .0131 .6281 .1931 .0978 INF .0290 .0640 .0518 .1147 INF .0214 .6080 .1430 .8135
15 (×100 ) .0000 .0000 .0000 .0012 .0048 .0040 .0029 .0058 .0106 .0130 .0038 .6601 .0266 .0137 .4225 .0075 .0153 .0151 .0413 INF .0103 .1661 .0095 .0335
16 (×100 ) .0000 .0000 .0000 .0249 .0063 .0024 .0038 .0061 .0136 .0139 .0023 .4428 .0217 .0093 .2376 .0104 .0261 .0154 .0145 .0556 .0112 .0586 .0179 .0098
20 (×100 ) .0000 .0000 .0000 .0002 .0100 .0045 .0058 .0099 .0197 .0160 .0389 .2714 .0625 .0190 .1268 .0131 .0140 .0152 .0171 .0428 .0115 .0171 .0066 .0059
21 (×10−1 ) .0000 .0000 .0000 .0017 .0760 .0408 .0317 .0000 .0901 .0854 .2074 .5621 .2829 .0689 INF .0857 .1411 .0622 .1900 .2160 .1213 .1480 .0525 .1313
22 (×10−1 ) .0000 .0000 .0000 .0777 .1154 .0215 .0043 .0616 .0533 .0696 .0712 .4278 INF .0748 .3273 .0723 .0260 .4899 .0913 .1250 .2968 .1564 .1506 .6375
24 (×102 ) .0000 .0000 .0000 .0010 .0368 .1151 .0300 .0444 .0684 .0972 .0239 INF .3539 .0780 INF .2266 .2926 .0667 .1199 INF .0510 .2606 .2433 INF
25 (×103 ) .0000 .0000 .0000 .0001 .0052 .0306 .0046 .0036 .0361 .0188 .0012 .1088 .0418 .0108 INF .1254 .2955 .0228 .1529 .1423 .0120 .0561 .0834 .0154
26 (×101 ) .0000 .0000 .0000 .0943 .0223 .0134 .0266 .0143 .0837 .1253 .0266 .8285 .0865 .0608 .3470 .0877 .1137 .0641 .1308 .2370 .0113 .3042 .0734 .1639
30 (×10−1 ) .0000 .0000 .0000 .0017 .0333 .0253 .0128 .0255 .0923 .0755 .0063 .1064 .1900 .0434 .5413 .0958 .0892 .1005 INF .3268 .0673 .0449 .0835 .0184
31 (×10−1 ) .0000 .0000 .0000 .0693 .0413 .0198 .0169 .0000 .0710 .1350 .0095 .3478 .2762 .0669 INF .1182 .1006 .0428 .1825 .3687 .1029 .0853 .1077 .0121
32 (×10−1 ) .0000 .0000 .0000 .0425 .0351 .0194 .0167 .0357 .0953 .0782 .0364 .6607 .3049 .0625 .7704 .1810 .1112 .0713 .0908 .6119 .0592 .0936 .1307 .0171
33 (×10−1 ) .0000 .0000 .0000 .0845 .0501 .0416 .0238 .0373 .1381 .1031 .0093 .6050 .2100 .0466 INF .0746 .0944 .0760 .0874 .3096 .0720 .0720 .1265 .0016
34 (×10−1 ) .0000 .0000 .0000 .0137 .0504 .0154 .0233 .0338 .0827 .1588 .0082 .6320 .3547 .0569 INF .1404 .1069 .1065 .0740 .5451 .0794 .0690 .1106 .0105
35 (×10−1 ) .0000 .0000 .0000 .0000 .0085 .0080 .0094 .0066 .0655 .1408 .0113 .3754 .0577 .0407 .4786 .2045 .7724 .1113 .7171 .1489 .0284 .3409 .3742 .1503
38 (×104 ) .0000 .0000 .0000 .0002 .0076 .0279 .0046 .0066 .0817 .1189 .0010 .0308 .1873 .0145 .0381 .0228 .0428 .5904 .0544 .0214 .1781 .0754 .0495 .2514
41 (×102 ) .0000 .0000 .0000 .0603 .0313 .0239 .0140 .0341 INF INF .0583 INF INF INF INF .0438 .0581 INF .0425 .8454 .8744 .5363 .0655 INF
54 (×101 ) .0000 .0000 .0000 .3159 .0489 .0368 .0090 .0310 .2831 .6400 .0184 .0352 INF OOM .0960 .0702 .0964 INF INF .0016 INF .0517 .0829 INF
57 (×100 ) .0000 .0000 .0000 .0005 .0124 .0029 .0029 .0000 .0146 .0207 .0094 .0784 .0222 .0099 .0523 .0148 .0239 .0179 .0166 .0042 .0133 .0171 .1242 INF
62 (×10−1 ) .0000 .0000 .0000 .0168 .0391 .1306 .0253 .0000 .0055 .1326 .0119 .0111 INF .0048 .0079 .0099 .0312 INF .0514 .0158 .0067 .0486 .0170 .0628
64 (×100 ) .0000 .0000 .0000 .0001 .0053 .0046 .0034 .0054 .0165 .0265 .0056 .0366 .0186 .0059 .5564 .0115 .0201 .0170 .0165 .0381 .0126 .0121 .0330 .0060
66 (×10−1 ) .0000 .0000 .0000 .0062 .0144 .0132 .0069 .0123 .0094 .0170 .0074 .0834 .1450 .0101 .1572 .0244 .0186 .0213 .0385 .2051 .0134 .0584 .0133 INF
69 (×10−1 ) .0000 .0000 .0000 .0000 .0012 .0021 .0019 .0006 .1349 .0500 .0003 .0205 .0078 .0373 .3199 .2092 .1762 .2472 .1978 .1903 .0007 .0669 .0361 .0392
70 (×10−1 ) .0000 .0000 .0000 .0002 .0072 .0066 .0051 .0120 .0714 .0388 .0022 .2971 .2853 .0324 INF .0561 .0316 .1229 .0150 .0985 .0133 .0968 .0251 .0152
71 (×10−1 ) .0000 .0000 .0000 .0000 .0032 .0216 .0131 .0025 .0594 .0381 .0143 .0588 .0589 .0711 .1000 .2085 .0954 .1284 .2391 .0744 .0579 .0844 .0597 .0166
73 (×100 ) .0000 .0000 .0000 .0197 .3876 .2585 .0833 .0000 .0357 .0164 .0130 .0093 .0569 .0319 .0026 .0539 .0288 .0225 .0253 .0063 .0514 .0208 .9433 INF
74 (×100 ) .0000 .0000 .0000 .0001 .0013 .0023 .0016 .0018 .0453 .0187 .0021 .0549 .0116 .0167 .2141 .0056 .0240 .0277 .0149 .0822 .0562 .0275 .0263 .1004
76 (×101 ) .0000 .0000 .0000 .0020 .0590 .0653 .0162 .0537 .0867 .0791 .0318 .1667 .1364 .1046 .3594 .1510 .1758 .1268 .1093 .3815 .0436 .1295 .1136 .8478
77 (×102 ) .0000 .0000 .0000 .4686 .0328 .0283 .0173 .0209 .0549 .0659 .0552 .6308 .1146 .0357 .6994 .0517 .0600 .0246 .0361 .0552 .0537 .0922 .0712 .0368
78 (×104 ) .0000 .0000 .0000 .0009 .0382 .0225 .0090 .0760 .0816 .0851 .0313 .5407 .6939 .0672 INF .0559 .1026 .0820 .1233 .0772 .0956 .1668 .1456 INF
35
79 (×104 ) .0000 .0000 .0000 .0007 .0131 .0065 .0018 .0931 .0120 .0264 .0042 .0857 .0880 .0119 .4144 .0099 .0205 .0210 .0130 .0026 .0151 .0490 .0152 INF
80 (×105 ) .0000 .0000 .0000 .0001 .0059 .0044 .0016 .0084 .0092 .0119 .0038 .0192 .1057 .0086 .0540 .0133 .0153 .0402 .0160 .0452 .0155 .0323 .0331 INF
81 (×104 ) .0000 .0000 .0000 .0005 .0238 .0083 .0027 .0123 .0934 .0552 .0160 .3037 .1830 .0196 INF .0379 .0422 .1126 .0494 .0182 .0305 .0621 .0187 INF
82 (×10−1 ) .0000 .0000 .0000 .0018 .0481 .0328 .0098 .0000 .0998 .1201 .1599 .3659 .1647 .0589 INF .1138 .0631 .1355 .0552 .1555 .0946 .0714 .0738 .0412
86 (×105 ) .0000 .0000 .0000 .0593 .0156 .0083 .0070 .0144 .0113 .0172 .0026 .0540 .0542 .0154 .0274 .0180 .0195 .0129 .0268 .0499 .0196 .0300 .0211 .0274
87 (×103 ) .0000 .0000 .0000 .0425 .0196 .0294 .0145 .0233 .0393 .0234 .0055 .0523 .1797 .0321 .0400 .0408 .0219 .0225 .0113 .0661 .0181 .0531 .0245 .0486
89 (×100 ) .0000 .0000 .0000 .1911 .0731 .0268 .0250 .0481 .1282 .1497 .0158 INF .4586 .1021 INF .0545 .1011 .0798 .1633 .8236 .0741 .0837 .0869 .1508
90 (×100 ) .0000 .0000 .0000 .0010 .0705 .1724 .0411 .0356 .0589 .0726 .0309 .9439 .8271 .0613 .1571 .0671 .0636 .1580 .1624 .1227 .0491 .3996 .2140 .0313
92 (×101 ) .0000 .0000 .0000 .0001 .0047 .0169 .0106 .0041 .0832 .0707 .0004 .1576 .0600 .0346 .4496 .1228 .0917 .0308 .0772 .0805 .0481 .0700 .0800 .0072
93 (×100 ) .0000 .0000 .0000 .0007 .0906 .0378 .0247 .0450 .0311 .0424 .0013 .2912 .4106 .0054 .2572 .0214 .0660 .0134 .1498 .0712 .1406 .0955 .2444 .0369
94 (×10−1 ) .0000 .0000 .0000 .0005 .0113 .0564 .0141 .0084 .2785 .2581 .0037 .2795 .1509 .2694 INF .1002 .3646 .0793 .0684 .3552 .0453 .1190 .1337 .0280
96 (×100 ) .0000 .0000 .0000 .0259 .0158 .0093 .0046 .0145 .0304 .0212 .0252 .1636 .0723 .0189 .1709 .0380 .0318 .0252 .9919 .0746 .0227 .0326 .0384 .0264
105 (×10−1 ) .0000 .0000 .0000 .0002 .0043 .0049 .0045 .0026 .3378 .0602 .0010 .0596 .0297 .0653 .3049 .0830 .0535 .3634 .0359 .1871 .0720 .0321 .4032 .0270
106 (×101 ) .0000 .0000 .0000 .0001 .0063 .0162 .0060 .0000 .0156 .0876 .0122 .0202 .0531 .0150 .2320 .0065 .0197 .0208 .0189 .0095 .0116 .0173 .0095 .0316
108 (×100 ) .0000 .0000 .0000 .0001 .0600 .1418 .0558 .0392 .0417 .1257 .0173 .1979 .5011 .0299 .1250 .0196 .1311 .2532 .3125 .2257 .0979 .1043 .0585 .1511
109 (×100 ) .0000 .0000 .0000 .0004 .0507 .1892 .0098 .0288 .1047 .1482 .0718 .1115 .7866 .0651 .1412 .1334 .0916 .0738 .4918 .0377 .0495 .2926 .1276 .0275
110 (×101 ) .0000 .0000 .0000 .0028 .0045 .0026 .0031 .0063 .0117 .0140 .0095 .1415 .0323 .0213 .7400 .0172 .0197 .0169 .0379 .0137 .0183 .0283 .0129 INF
113 (×101 ) .0000 .0000 .0000 .0001 .0203 .0086 .0137 .0158 .1060 .0497 .0163 INF .1431 .0263 .4573 .1349 OOM .3555 INF .2364 .0577 INF OOM INF
114 (×10−2 ) .0000 .0000 .0000 .5501 .0211 .2784 .0331 .0037 .1362 .2501 .0235 .0158 .2011 .0496 .0262 .0707 .1112 .4447 .5152 .0245 .0327 .1914 .0222 .0504
116 (×10−1 ) .0000 .0000 .0000 .6470 .0229 .0361 .0186 .0233 .0863 .0568 .0016 .2914 .2042 .0444 INF .0715 .0716 .1137 .2790 .3750 .0352 .1243 .2831 .0814
118 (×10−1 ) .0000 .0000 .0000 .1224 .0850 .0536 .0438 .0507 .1088 .0958 .0019 .3600 .4122 .0617 .4055 .0885 .0576 .1061 .1927 .1295 .1079 .1676 .1351 .0663
119 (×10−1 ) .0000 .0000 .0000 .0008 .0557 .0263 .0208 .0502 .1091 .0573 .0113 .0311 .2893 .1200 .0761 .0478 .0850 .4064 .0595 .0452 .0649 .0788 .0924 .0190
125 (×100 ) .0000 .0000 .0000 .0925 .0413 .0240 .0272 .0000 .1325 .0319 .0401 .0824 .4467 .1915 INF .0720 .0548 .0748 .1401 .1104 .1320 .2008 .0660 .0165
130 (×10−1 ) .0000 .0000 .0000 .0004 .0258 .0100 .0113 .0000 .0241 .0347 .0047 .1035 .7527 .0397 .2299 .0218 .0269 .0188 .0561 .0482 .0235 .0947 .0585 .2477
131 (×102 ) .0000 .0000 .0000 .0423 .0441 .0459 .0404 .0498 .0884 .0605 .0348 .0524 .3183 .1105 .0363 .1142 .1236 .1058 .1304 .0634 .0875 .0791 .1499 .0481
132 (×100 ) .0000 .0000 .0000 .1614 .1187 .1580 .0395 .0537 .0667 .2924 .0161 .0191 .7523 .0498 .3922 .0970 .0845 .1857 .0846 .0662 .0345 .1087 .2974 .0343
133 (×100 ) .0000 .0000 .0000 .0003 .1086 .0478 .0160 .0272 .0360 .0486 .0121 .0096 .6151 .0576 .0529 .1763 .0249 .0064 .0385 INF .0777 .1036 .1069 .0294
135 (×104 ) .0000 .0000 .0000 .0218 .0108 .0191 .0018 .0000 .0384 .0214 .0046 .0432 .0478 .0154 .0122 .0469 .0396 .0528 .0358 .0104 .0371 .0219 .1055 .0079
138 (×10−1 ) .0000 .0000 .0000 .0002 .0068 .0070 .0047 .0114 .0704 .0479 .0484 .1715 .1224 .0214 .1115 .0191 .0208 .0207 .0369 .0710 .0147 .0403 .0170 .0647
140 (×10−2 ) .0000 .0000 .0000 .0377 .0522 .0496 .0226 .0451 .0668 .0872 .0132 .1300 .1201 .0660 INF .0877 .1056 .0612 .1161 .3897 .0391 .3675 .0704 .4497
141 (×10−2 ) .0000 .0000 .0000 .0008 .0123 .0054 .0097 .0183 .0239 .0447 OOM .9598 .1518 .0268 .8714 .0334 .0392 .0392 .0387 .1317 .0209 .1150 .0209 .0135
144 (×101 ) .0000 .0000 .0000 .1437 .0091 .0175 .0103 .0109 .0900 .0667 .0046 .0128 .3021 .0464 INF .1175 OOM .0322 .3627 OOM .0290 .1171 .1691 INF
148 (×100 ) .0000 .0000 .0000 .0019 .0290 .0178 .0107 .0000 .8221 INF .0108 .6779 .0888 .5751 .0710 .5308 .8714 INF INF .0486 .0846 .1572 INF .1312
149 (×10−1 ) .0000 .0000 .0000 .0002 .0103 .0216 .0102 .0272 .0135 .0312 .0014 .1425 .0731 .0225 INF .0244 .0344 .0172 .0148 .0573 .1105 .0333 .0341 .0047
152 (×10−1 ) .0000 .0000 .0000 .0002 .0286 .0118 .0057 .0220 .0680 .0592 .0018 .1797 .0205 .0179 INF .0385 .1034 .0809 .0334 .0277 .0742 .1482 .0859 INF
153 (×101 ) .0000 .0000 .0000 .0003 .0119 .0113 .0070 .0000 .0206 .0951 .0038 .0257 .0975 .0284 .2060 .0331 .0495 .0364 .0700 .0461 .0199 .0250 .0437 .0136
155 (×100 ) .0000 .0000 .0000 .0025 .0695 .0396 .0063 .0116 .0303 .0504 .0099 .3332 .4141 .0528 .5662 .0465 .0492 .0560 .0876 .1072 .0304 .2382 .0489 .0221
156 (×100 ) .0000 .0000 .0000 .0016 .0315 .0142 .0097 .0158 .0330 .0268 .0160 .1427 .2301 .0210 .4520 .0523 .0257 .0877 .0721 .2560 .0443 .2556 .0296 .1674
158 (×10−1 ) .0000 .0000 .0000 .0002 .0041 .0028 .0030 .0054 .0930 .0178 .0114 .1123 .0787 .0205 .4223 .0375 .0583 .0507 .0902 .0668 .0097 .0593 .1780 INF
161 (×100 ) .0000 .0000 .0000 .0103 .0108 .0131 .0074 .0147 .0332 .0266 .0206 .0529 .0729 .0497 .1513 .0464 .0861 .0584 .0290 .0912 .0068 .0517 .0650 .0144
162 (×10−1 ) .0000 .0000 .0000 .0006 .0186 .0179 .0126 .0000 .0414 .0281 .0168 .0306 .1216 .0216 .0080 .0365 .0391 .0499 .0179 .0054 .0214 .0300 .0414 .0096
165 (×10−4 ) .0000 .0000 .0000 .0410 .0295 .0413 .0155 .0342 .0654 .0388 .0030 INF .1111 .0324 .5303 .0585 .0251 .0527 .0657 .5437 .0391 .1078 .0487 .2730
172 (×10−3 ) .0000 .0000 .0000 .0198 .0223 .0133 .0104 .0148 .0092 .0115 .0408 .0884 .0496 .0108 .3201 .0098 .0111 .0101 .0186 .1145 .0138 .0577 .0105 .0107
176 (×10−1 ) .0000 .0000 .0000 .0009 .0128 .0198 .0062 .0215 .0512 .0656 .0078 .3248 .0876 .0653 INF .0698 .0279 .0471 .0465 .0830 .0114 .0687 .0394 .0482
178 (×10−1 ) .0000 .0000 .0000 .0026 .0312 .0110 .0076 .0409 .0367 .0276 .0055 .1192 .0906 .0362 .1125 .0192 .0280 .0128 .1052 .3451 .0430 .2293 .0240 .1707
179 (×10−1 ) .0000 .0000 .0000 .0002 .0128 .0336 .0072 .0165 .0741 .0351 .0035 .0711 .0736 .0045 .0167 .0093 .0224 .0159 .0365 .0344 .0348 .0354 .0192 .0138
181 (×100 ) .0000 .0000 .0000 .0004 .0070 .0041 .0019 .0000 .3515 INF .0112 .1001 .1050 .0167 .0738 .1768 .7915 INF INF .0438 .0297 .1531 .2921 .0074
184 (×103 ) .0000 .0000 .0000 .0011 .0130 .0600 .0135 .0000 .0538 .0467 .0757 .3494 .4140 .0472 .8506 .0703 .0717 .2147 .8217 .6573 .0522 INF .0429 .1928
188 (×104 ) .0000 .0000 .0000 .0475 .0308 .0186 .0076 .0214 .0401 .0258 .0137 .0395 .0888 .0232 .1030 .0284 .0519 .0229 .0478 .1123 .0524 .0541 .0464 .0192
189 (×104 ) .0000 .0000 .0000 .3976 .0148 .0185 .0060 .0122 .0319 .0223 .0161 .0783 .0668 .0309 INF .0425 .0510 .0414 .0441 .0746 .0686 .0273 .0418 .2057
190 (×105 ) .0000 .0000 .0000 .1014 .0451 .0130 .0165 .0071 .0199 .0197 .0116 .0901 .1417 .0345 .7826 .0255 .0579 .0178 .0660 .0755 .0153 .0334 .0773 .0071
191 (×104 ) .0000 .0000 .0000 .0112 .0832 .0194 .0157 .0519 .0778 .0390 .0512 .4721 .1894 .0410 .3048 .0603 .0687 .0494 .0307 .1381 .0375 .0514 .0608 .0273
192 (×106 ) .0000 .0000 .0000 .0004 .0131 .0553 .0129 .0376 .0128 .0215 .0139 .0712 .1410 .0217 .0761 .0175 .0141 .0057 .0324 INF .0384 .1113 .0378 .0369
202 (×10−2 ) .0000 .0000 .0000 .0021 .1794 .0531 .0512 .0000 .1067 .1064 .0398 .3845 .4509 .1592 INF .1299 .1704 .0668 .1432 .3329 .0900 .2891 .0827 .0399
205 (×10−2 ) .0000 .0000 .0000 .0221 INF .0035 .0000 .0000 .8136 .6531 .0027 .3815 .8979 .2126 INF .0089 .3845 .0198 .0178 INF .1435 INF .0080 .1331
212 (×100 ) .0000 .0000 .0000 .0266 .1136 .0158 .0052 .0000 .0307 .2564 .0101 INF .3554 .0125 INF .0186 .0316 .0414 .0515 .1902 .0279 .2750 .0097 .0173
223 (×10−1 ) .0000 .0000 .0000 .0107 .3991 .0414 .0493 .0000 .1276 .1795 .0024 INF .4863 .4431 INF .0873 .1043 .0537 .1124 INF .1692 .2673 .0582 .0202
233 (×10−1 ) .0000 .0000 .0000 .0000 .0164 .0391 .0076 .0082 .0526 .1550 .0025 .0052 .0931 .0452 .0209 .0222 .0577 .0296 .0793 .0062 .0816 .0155 .0106 .0102
241 (×101 ) .0000 .0000 .0000 .0072 .0099 .0054 .0033 .0127 .0134 .0205 .0098 .0471 .0806 .0106 .1579 .0081 .0157 .0072 .0303 .1220 .0170 .0748 .0058 .1159
242 (×101 ) .0000 .0000 .0000 .0060 .0077 .0065 .0030 .0062 .0191 .0360 .0077 .0647 .0871 .0099 INF .0105 .0104 .0128 .0244 .1384 .5394 .0751 .0096 .0379
244 (×10−3 ) .0000 .0000 .0000 .0029 .0648 .0538 .0196 .0472 .2824 .1451 .1486 .0687 INF .1526 .0083 .7134 .0854 .1267 .3188 .0501 .1668 .8579 .2799 .0701
245 (×100 ) .0000 .0000 .0000 .0003 .0076 .0093 .0050 .0000 .0138 .0292 OOM .1234 .0415 .0243 .0305 .0080 .0123 .0121 .0511 .0373 .0095 .0766 .0201 .0112
247 (×100 ) .0000 .0000 .0000 .0956 .0344 .0265 .0189 .0097 .0355 .0223 .0291 .0166 .1349 .0261 .0203 .0326 .0243 .0356 .0432 .0194 .0464 .0483 .0167 .0292
248 (×100 ) .0000 .0000 .0000 .0589 .0241 .0137 .0075 .0108 .0133 .0549 .0119 .0258 .0574 .0219 .1209 .0476 .0305 .0213 .0200 .0215 .0316 .0745 .0392 .0028
252 (×100 ) .0000 .0000 .0000 .0055 .0082 .0046 .0034 .0064 .0173 .0207 .0023 .2236 .0323 .0220 INF .0281 .0213 .0282 .0519 .1366 .0235 .0669 .0270 .0033
257 (×100 ) .0000 .0000 .0000 .0240 .0152 .0157 .0041 .0059 .0243 .0295 .0142 .0014 .0307 .0109 .0020 .0291 .0243 .0048 .0209 .0024 .0121 .0850 .0182 .0087
258 (×10−1 ) .0000 .0000 .0000 .0000 .0077 .0199 .0046 .0000 .0988 .0195 .0013 .0816 .0173 .0075 .2153 .1725 .2203 .0259 .0964 .1313 .0129 .0275 .1656 .0112
259 (×10−1 ) .0000 .0000 .0000 .0000 .0074 .0058 .0044 .0033 .1176 .0459 .0015 .0415 .0450 .0308 .0583 .0442 .0490 .1565 .0773 .1394 .0162 .0469 .0413 .0205
261 (×102 ) .0000 .0000 .0000 .0007 .0127 .0390 .0043 .0055 .0437 .0202 .0042 .0189 .0498 .0030 INF .0100 .0123 .0110 .0242 .0123 .0060 .0674 .0145 .0030
262 (×10−1 ) .0000 .0000 .0000 .0311 .0231 .0092 .0058 .0000 .0114 .0154 .0019 .0901 .3780 .0152 .0360 .0181 .0199 .0345 .0075 .0403 .0165 .0292 .0272 .0094
267 (×100 ) .0000 .0000 .0000 .0303 .0382 .0406 .0072 .0410 .0694 .1502 .0084 .1408 .2099 .1826 INF .0398 .0622 .0315 .0626 INF .0956 .1714 .1196 .3284
269 (×100 ) .0000 .0000 .0000 .0013 .0389 .0284 .0212 .0436 .0157 .0245 .0145 .3039 .0838 .0383 INF .0187 .0173 .0278 .0418 .1514 .4194 .1626 .0200 .0072
270 (×101 ) .0000 .0000 .0000 .3528 .2944 .1112 .0392 .0027 .0705 .0292 .0043 .0094 .2915 .0660 .0146 .0261 .1024 .1498 .2295 .0113 .1541 .0544 .3814 .0589
271 (×10−2 ) .0000 .0000 .0000 .1687 .0842 .1077 .0412 .0454 .2877 .0445 .0039 .0827 .3085 .2346 .0403 .2832 .1067 .3434 .0935 .0447 .0992 .0568 .1480 .0315
272 (×100 ) .0000 .0000 .0000 .0000 .0049 .0041 .0050 .0043 .0556 .0204 .0059 .0403 .0281 .0137 .1394 .0404 .0175 .0278 .0187 .0646 .0141 .0227 .0696 .0032
280 (×10−2 ) .0000 .0000 .0000 .0001 .0034 .0047 .0047 .0061 .0243 .0252 .0185 .0190 .0438 .0286 .0155 .0334 OOM INF .0352 .0150 .0049 .0841 .0240 INF
281 (×10−1 ) .0000 .0000 .0000 .0034 .1074 .0926 .0818 .0646 .1405 .0663 .0764 INF INF .0633 INF .0881 .1321 .0874 .1797 INF .1406 .9235 .1950 .0025
286 (×101 ) .0000 .0000 .0000 .1182 .0414 .0486 .0253 .0455 .1755 INF .0676 .0484 .4025 .0783 .2444 .1383 .2026 .3007 .0997 .0461 .0876 .1051 .1130 INF
293 (×100 ) .0000 .0000 .0000 .0089 .0228 .0157 .0083 .0763 .0177 INF .0257 INF .2158 .0171 INF .0224 .0224 .0096 .1555 .5568 .0239 .4566 .0185 .1025
295 (×100 ) .0000 .0000 .0000 .0006 .0151 .0082 .0067 .0137 .0146 .0235 .0079 .2805 .0848 .0243 INF .0160 .0211 .0191 .0214 .0431 .0287 .0385 .0150 .0327
297 (×10−1 ) .0000 .0000 .0000 .0012 .0420 .0384 .0204 .0460 .1122 .0570 OOM .0408 .1228 .0650 .1081 .0688 .0845 .0933 .0554 .1001 .0545 .0558 .1039 .0578
36
D.2 Statistically Significant Performance Differences
Based on Table 1 in the main paper, the differences between many methods are not readily apparent.
Therefore, we use statistical tests to measure the performance differences between different methods.
For any two methods A and B and a dataset D, we perform a t-test on the results of 15 seeds for
each method, which helps determine whether the performance differences between each algorithm
are significant or not with 95% confidence interval. If the p-value is greater than 0.05, the methods
are considered tied. Otherwise, if the null hypothesis is rejected, we consider the performance
difference between A and B on dataset D to be significant. In this case, we use the mean value of the
corresponding metric to determine that one method is significantly better than the other.
In this section, we select the two best-performing tree-based methods (XGBoost and CatBoost) and
the two best-performing DNN-based methods (FT-T and TabR) as the anchor mehtods, and compare
their results against all other methods in the benchmark across all datasets. Based on the comparisons
across 300 datasets, we use the win/tie/lose frequency to calculate the win rate, tie rate, and lose rate
for each pair of methods. These results are recorded in Table 19, Table 21, Table 20, and Table 22,
which correspond to the results of t-tests conducted on 101 binary classification datasets, 80 multi-
class classification datasets, 119 regression datasets, and all 300 datasets in the full benchmark,
respectively.
The results in Table 22 show that CatBoost has a tie rate of 0.3 when compared to XGBoost and 0.26
when compared to LightGBM, demonstrating the high similarity among Gradient Boosting Decision
Trees (GBDTs). CatBoost also shows outstanding performance, achieving a win rate of over 0.5 in
comparisons with most methods. However, its win rate against TabR is only 0.4358, highlighting the
significant potential of Neighborhood-based methods for tabular prediction tasks. This indicates that
while CatBoost excels broadly, Neighborhood-based methods like TabR are highly competitive and
can outperform tree-based methods in certain scenarios.
Additionally, Figure 7, Figure 8, Figure 9, and Figure 10 illustrate the significance of performance
differences between the selected methods and other methods across 300 datasets. The heatmaps
display these differences, with darker squares indicating smaller p-values. Red squares signify that
the selected method performs better, while blue ones indicate superior performance by the compared
method.
Comparing Figure 9 and Figure 10, we find that FT-T and TabR perform differently against GBDTs.
Specifically, TabR’s heatmap shows less blue in the corresponding sections, indicating that TabR
has fewer cases where it performs worse compared to GBDTs, highlighting its relatively stronger
performance in these scenarios. This suggests that Neighborhood-based approaches can be beneficial
for deep learning on tabular data.
37
Table 19: The win/tie/lose rates of Method A (CatBoost, XGBoost, FT-T, and TabR) against Method
B across 101 binary classification datasets.
CatBoost XGBoost FT-T TabR
Method B Win Tie Lose Win Tie Lose Win Tie Lose Win Tie Lose
Dummy .9625 .0250 .0125 .9750 .0125 .0125 .9750 .0125 .0125 .9744 .0128 .0128
LR .7750 .0500 .1750 .6875 .1000 .2125 .7125 .1500 .1375 .8077 .0897 .1026
NCM .9750 .0125 .0125 .9750 .0000 .0250 .9375 .0375 .0250 .9744 .0128 .0128
Naive Bayes .9750 .0125 .0125 .9500 .0375 .0125 .9250 .0500 .0250 .9359 .0385 .0256
KNN .7625 .0500 .1875 .6625 .0750 .2625 .6750 .1250 .2000 .7821 .1154 .1026
SVM .8125 .0750 .1125 .7875 .0500 .1625 .7750 .1500 .0750 .8590 .0769 .0641
XGBoost .4625 .3625 .1750 .0000 1.000 .0000 .4000 .2500 .3500 .5256 .1795 .2949
CatBoost .0000 1.000 .0000 .1750 .3625 .4625 .2875 .2375 .4750 .4231 .2692 .3077
Random Forest .6750 .1375 .1875 .5625 .2750 .1625 .5250 .1750 .3000 .5513 .1923 .2564
LightGBM .4125 .2625 .3250 .2875 .3125 .4000 .3875 .2125 .4000 .4615 .2692 .2692
TabPFN .6667 .1667 .1667 .6500 .1167 .2333 .5167 .3167 .1667 .6610 .1695 .1695
MLP .5500 .1875 .2625 .4750 .2250 .3000 .4125 .3750 .2125 .6154 .2949 .0897
ResNet .5000 .1750 .3250 .4500 .1500 .4000 .3625 .3500 .2875 .5513 .3205 .1282
NODE .8462 .1026 .0513 .7692 .1282 .1026 .7051 .2308 .0641 .7895 .1579 .0526
SwitchTab .8875 .0500 .0625 .8125 .1000 .0875 .8000 .1625 .0375 .8846 .0897 .0256
TabNet .8875 .0875 .0250 .8875 .0750 .0375 .9375 .0500 .0125 .9231 .0641 .0128
TabCaps .6375 .1625 .2000 .5250 .2250 .2500 .5250 .4000 .0750 .7179 .1923 .0897
TANGOS .6250 .1625 .2125 .5250 .2000 .2750 .5250 .3250 .1500 .6667 .2179 .1154
DANets .6125 .1750 .2125 .5625 .2250 .2125 .5625 .3125 .1250 .7692 .2051 .0256
FT-T .4750 .2375 .2875 .3500 .2500 .4000 .0000 1.000 .0000 .4872 .3462 .1667
AutoInt .5875 .1375 .2750 .5375 .1625 .3000 .5000 .3750 .1250 .6154 .2821 .1026
DCNv2 .5375 .1875 .2750 .4375 .2000 .3625 .4000 .4500 .1500 .5641 .2949 .1410
SNN .5625 .1625 .2750 .4750 .2250 .3000 .4125 .4250 .1625 .6538 .2436 .1026
TabTransformer .7468 .1266 .1266 .6456 .1772 .1772 .7089 .2025 .0886 .7662 .1299 .1039
PTaRL .6250 .1750 .2000 .5750 .1750 .2500 .5375 .4250 .0375 .7821 .1410 .0769
GrowNet .7600 .1733 .0667 .7867 .1200 .0933 .7733 .1333 .0933 .8514 .1081 .0405
TabR .3077 .2692 .4231 .2949 .1795 .5256 .1667 .3462 .4872 .0000 1.000 .0000
Table 20: The win/tie/lose rates of Method A (CatBoost, XGBoost, FT-T, and TabR) against Method
B across 119 regression datasets.
CatBoost XGBoost FT-T TabR
Method B Win Tie Lose Win Tie Lose Win Tie Lose Win Tie Lose
Dummy .9580 .0084 .0336 .9664 .0084 .0252 .8908 .0672 .0420 .8814 .0339 .0847
KNN .8655 .0504 .0840 .8403 .0336 .1261 .6471 .0756 .2773 .6441 .0508 .3051
SVM .9244 .0336 .0420 .9328 .0336 .0336 .8235 .0504 .1261 .7712 .0508 .1780
XGBoost .6218 .1765 .2017 .0000 1.000 .0000 .3109 .2017 .4874 .3814 .1186 .5000
CatBoost .0000 1.000 .0000 .2017 .1765 .6218 .2353 .1681 .5966 .3305 .1017 .5678
Random Forest .7143 .1092 .1765 .6975 .0924 .2101 .4538 .1345 .4118 .4915 .0932 .4153
LightGBM .6050 .1765 .2185 .3529 .2269 .4202 .3277 .1345 .5378 .3729 .0593 .5678
MLP .6891 .1345 .1765 .6050 .1261 .2689 .5546 .2521 .1933 .5763 .2458 .1780
ResNet .6975 .1008 .2017 .6218 .0924 .2857 .5294 .2185 .2521 .5424 .2373 .2203
NODE .7500 .0517 .1983 .6983 .0603 .2414 .4828 .2155 .3017 .4957 .1826 .3217
SwitchTab .9832 .0000 .0168 .9832 .0000 .0168 .8908 .0420 .0672 .8475 .0678 .0847
TabNet .9328 .0252 .0420 .8908 .0756 .0336 .7815 .1176 .1008 .7542 .1017 .1441
TANGOS .7373 .0847 .1780 .6186 .0763 .3051 .5678 .2203 .2119 .5983 .1709 .2308
DANets .9160 .0672 .0168 .9160 .0672 .0168 .8151 .1176 .0672 .7881 .1525 .0593
FT-T .5966 .1681 .2353 .4874 .2017 .3109 .0000 1.000 .0000 .4153 .2712 .3136
AutoInt .7069 .0948 .1983 .5776 .0948 .3276 .5086 .3621 .1293 .5259 .2586 .2155
DCNv2 .6891 .1429 .1681 .6134 .1008 .2857 .5546 .2437 .2017 .5085 .2373 .2542
SNN .7647 .1008 .1345 .7143 .0756 .2101 .6303 .1933 .1765 .6017 .2542 .1441
TabTransformer .9746 .0000 .0254 .9746 .0000 .0254 .8814 .0508 .0678 .8547 .0684 .0769
PTaRL .6807 .0756 .2437 .5294 .1597 .3109 .4538 .1765 .3697 .4237 .1949 .3814
GrowNet .9412 .0420 .0168 .9160 .0588 .0252 .7983 .1261 .0756 .7712 .1017 .1271
TabR .5678 .1017 .3305 .5000 .1186 .3814 .3136 .2712 .4153 .0000 1.000 .0000
DNNR .8824 .0588 .0588 .8403 .0588 .1008 .7899 .0924 .1176 .8220 .0593 .1186
38
Table 21: The win/tie/lose rates of Method A (CatBoost, XGBoost, FT-T, and TabR) against Method
B across 80 multi-class classification datasets.
CatBoost XGBoost FT-T TabR
Method B Win Tie Lose Win Tie Lose Win Tie Lose Win Tie Lose
Dummy .9307 .0396 .0297 .9307 .0297 .0396 .8812 .0792 .0396 .8900 .0600 .0500
LR .6733 .0990 .2277 .6931 .1089 .1980 .6139 .1584 .2277 .6200 .2000 .1800
NCM .9802 .0000 .0198 .9703 .0099 .0198 .9703 .0099 .0198 .9600 .0200 .0200
Naive Bayes .9505 .0000 .0495 .9406 .0099 .0495 .9109 .0495 .0396 .9200 .0300 .0500
KNN .7921 .0792 .1287 .7921 .0792 .1287 .7426 .0693 .1881 .7600 .0800 .1600
SVM .6733 .1485 .1782 .6931 .1089 .1980 .6238 .1782 .1980 .6300 .2600 .1100
XGBoost .3960 .3960 .2079 .0000 1.000 .0000 .2772 .3366 .3861 .3700 .2700 .3600
CatBoost .0000 1.000 .0000 .2079 .3960 .3960 .2178 .3564 .4257 .3000 .3200 .3800
Random Forest .5842 .2673 .1485 .5644 .2475 .1881 .4752 .2079 .3168 .5000 .2200 .2800
LightGBM .3762 .3564 .2673 .2673 .3960 .3366 .2970 .2871 .4158 .3300 .2600 .4100
TabPFN .6327 .1224 .2449 .6020 .1224 .2755 .5204 .1735 .3061 .5052 .2474 .2474
MLP .5743 .2277 .1980 .5149 .2079 .2772 .4752 .2871 .2376 .5200 .3600 .1200
ResNet .6238 .1881 .1881 .5842 .1584 .2574 .4950 .3069 .1980 .4800 .4100 .1100
NODE .7723 .1386 .0891 .7426 .1188 .1386 .6436 .2178 .1386 .6300 .1900 .1800
SwitchTab .8119 .1782 .0099 .8020 .1584 .0396 .7723 .1782 .0495 .7900 .1900 .0200
TabNet .8614 .1188 .0198 .8416 .1188 .0396 .8119 .1782 .0099 .8600 .1300 .0100
TabCaps .5941 .2178 .1881 .5545 .2079 .2376 .4851 .3564 .1584 .5600 .3100 .1300
TANGOS .6238 .2574 .1188 .6139 .2079 .1782 .5050 .3366 .1584 .5500 .3500 .1000
DANets .6139 .2178 .1683 .5842 .1584 .2574 .4950 .3663 .1386 .5200 .3600 .1200
FT-T .4257 .3564 .2178 .3861 .3366 .2772 .0000 1.000 .0000 .3800 .3900 .2300
AutoInt .5300 .2800 .1900 .4000 .3600 .2400 .3800 .4400 .1800 .4444 .3737 .1818
DCNv2 .6040 .1782 .2178 .5347 .1782 .2871 .4455 .3267 .2277 .5100 .3300 .1600
SNN .6040 .1980 .1980 .5050 .2178 .2772 .4554 .3663 .1782 .5500 .2700 .1800
TabTransformer .6337 .1782 .1881 .6337 .1188 .2475 .5446 .2277 .2277 .6100 .2400 .1500
PTaRL .5347 .3069 .1584 .5149 .2772 .2079 .4059 .4554 .1386 .5200 .3600 .1200
GrowNet .7228 .1485 .1287 .6733 .1980 .1287 .6832 .2475 .0693 .7300 .1900 .0800
TabR .3800 .3200 .3000 .3600 .2700 .3700 .2300 .3900 .3800 .0000 1.000 .0000
Table 22: The win/tie/lose rates of Method A (CatBoost, XGBoost, FT-T, and TabR) against Method
B across 300 datasets.
CatBoost XGBoost FT-T TabR
Method B Win Tie Lose Win Tie Lose Win Tie Lose Win Tie Lose
Dummy .9500 .0233 .0267 .9567 .0167 .0267 .9100 .0567 .0333 .9088 .0372 .0541
KNN .8133 .0600 .1267 .7767 .0600 .1633 .6867 .0867 .2267 .7196 .0777 .2027
SVM .8100 .0833 .1067 .8133 .0633 .1233 .7433 .1200 .1367 .7466 .1284 .1250
XGBoost .5033 .3000 .1967 .0000 1.000 .0000 .3233 .2600 .4167 .4155 .1858 .3986
CatBoost .0000 1.000 .0000 .1967 .3000 .5033 .2433 .2500 .5067 .3446 .2196 .4358
Random Forest .6600 .1700 .1700 .6167 .1933 .1900 .4800 .1700 .3500 .5101 .1622 .3277
LightGBM .4767 .2600 .2633 .3067 .3067 .3867 .3333 .2067 .4600 .3818 .1824 .4358
MLP .6133 .1800 .2067 .5400 .1800 .2800 .4900 .2967 .2133 .5676 .2973 .1351
ResNet .6200 .1500 .2300 .5633 .1300 .3067 .4733 .2833 .2433 .5236 .3176 .1588
NODE .7831 .0949 .1220 .7322 .0983 .1695 .5966 .2203 .1831 .6186 .1787 .2027
SwitchTab .9000 .0733 .0267 .8767 .0800 .0433 .8267 .1200 .0533 .8378 .1149 .0473
TabNet .8967 .0733 .0300 .8733 .0900 .0367 .8333 .1200 .0467 .8345 .1014 .0642
TANGOS .6689 .1639 .1672 .5920 .1538 .2542 .5351 .2876 .1773 .6000 .2441 .1559
DANets .7333 .1467 .1200 .7100 .1400 .1500 .6400 .2533 .1067 .6926 .2365 .0709
FT-T .5067 .2500 .2433 .4167 .2600 .3233 .0000 1.000 .0000 .4223 .3311 .2466
AutoInt .6149 .1689 .2162 .5068 .2027 .2905 .4628 .3919 .1453 .5222 .3038 .1741
DCNv2 .6200 .1667 .2133 .5400 .1533 .3067 .4767 .3267 .1967 .5236 .2838 .1926
SNN .6567 .1500 .1933 .5800 .1633 .2567 .5133 .3133 .1733 .5980 .2568 .1453
TabTransformer .7987 .0940 .1074 .7718 .0872 .1409 .7215 .1510 .1275 .7483 .1429 .1088
PTaRL .6167 .1800 .2033 .5367 .2033 .2600 .4600 .3367 .2033 .5507 .2365 .2128
GrowNet .8203 .1119 .0678 .8000 .1220 .0780 .7525 .1695 .0780 .7774 .1336 .0890
TabR .4358 .2196 .3446 .3986 .1858 .4155 .2466 .3311 .4223 .0000 1.000 .0000
39
1
21
41
61
81
101
121
Dataset_ID
141
161
181
201
221
241
261
281
40
1
21
41
61
81
101
121
Dataset_ID
141
161
181
201
221
241
261
281
41
1
21
41
61
81
101
121
Dataset_ID
141
161
181
201
221
241
261
281
42
1
21
41
61
81
101
121
Dataset_ID
141
161
181
201
221
241
261
281
43