greatjearning

Le n i ng Li fe

Curriculum Overview
 Program Overview
Recommended if you prefer learning online 72+ hours of online mento
12 Month Course
150+ hours of online resources
Individual doubt-solving with expert mentors Personalised feedback fr
Dedicated Online groups to interact with your mentor

Recommended if you prefer learning in the classroom 250+ hours of c

12 Month Course
150+ hours online resources
Live faculty interactions and doubt solving Interact and network with 5
Lab work with faculty guidance

ONLINE+MENTORSHIP MODE
WEEKEND CLASSROOM MODE
 TEXAS
The University of Texas at Austin

Overview PGP AIML

view
ng online 72+ hours of online mentorship

t mentors Personalised feedback from industry experts

ct with your mentor

ng in the classroom 250+ hours of classroom sessions

solving Interact and network with 50+ peers

PGP AIML
 AIML – Course Curriculum
MODULE 1
INTRODUCTION TO PYTHON
Installation of Python
NumPy

Basics of NumPy array Broadcasting an array Matrix

indexing
Pandas
Series
Indexing elements of a series
Dictionaries Data frame Different ways of
indexing in a data frame
Conditional indexing in a data frame

Seaborn for data visualization

Univariate analysis
– Dist plot, count
plot, Boxplot,

Case Study
MODULE 2
STATISTICAL LEARNING
Introduction to Statistical analysis
se Curriculum
O PYTHON

Selection techniques Saving and loading

arrays
Dropna and fillna
Groupby, merging similar to SQL logic
Filtering, sorting and indexing
Loops and functions
Saving and loading a csv, excel file

ta visualization
Bivariate analysis – Pair plot, Reg plot, Joint
plot, Point plot, Factor plot, Strip plot, Swarm
plot
RNING
al analysis
Why statistics?
Popular statistical methods
in the field of AI & ML
Types of statistics –
Descriptive & Inferential
Descriptive Statistics
Histogram
Frequency distribution
Measures of central
tendency (3M) – Mean,
median, mode
Measures of dispersion
– Range, IQR, SD,
variance & coefficient of
variation
Introduction to
probability
Basic probability: Definition
and examples
Mutually exclusive events
Independent events
Rules for computing
probability
Conditional probability
Common terminologies in Statistics
Data sources – Primary and Secondary
Types of data – Qualitative and Quantitative
Empirical and Chebyshev rule Five number
summary,
QQ plots, Quantile plot, scatterplot.
Visualization – Scatter plot, correlation analysis
Baye’s theorem
Applications of Bayes theorem in real life
scenario
Probability distributions Binomial distribution
Poisson distribution
Normal distribution
 Hypothesis testing & scores Concept of sampling distributi
theorem
Formulation of hypothesis – Null and Alternate Type I and II error
Hypothesis testing roadmap
Type of Hypothesis – One tail vs Two tailed test Confidence Inter
Single sample - Z test Single sample - T test
Two sample test
Paired T test
Z test of proportion
Chi-square test of variance F-ratio test of variance ANOVA

MODULE 3
SUPERVISED LEARNING
Introduction to machine learning

What is machine
learning?
When is machine
learning useful?
Application of
machine learning in
different domains
& scores Concept of sampling distribution Central Limit
s – Null and Alternate Type I and II error
ap
e tail vs Two tailed test Confidence Interval
ngle sample - T test

ce F-ratio test of variance ANOVA

RNING
learning

Categories of machine learning

Supervised learning – concepts, preparing
data, training & testing
Unsupervised learning – concepts and usage
Supervised ML: Linear Regression
Linear relationship R2 and adjusted R2 Multip
between two variables Advantages & disadvantag
Measures of regression
association: Pearson’s How to treat independent c
coefficient Case study
Types of correlation
Best fit line – SSE,
SSR and SST
Standard error, MAE
and RMSE

Logistic regression
Logistic function – Sigmoid
curve
Log loss Odds ratio
Advantages & disadvantages of
logistic regression

K-NN classification
Distance measures
Euclidean, Manhattan,
Minkowski distance measures
Mahalanobis distance, Cosine
similarity
Radius neighbour & KD Tree
classifier

Naïve Bayesian
classifiers Conditional
probability Independent events
Baye’s theorem
Assumptions of Naïve Baye’s
classifier
ear Regression
R2 and adjusted R2 Multiple linear regression
Advantages & disadvantages of Linear
regression
How to treat independent categorical variable
Case study

Multiple dimensions Setting up threshold

Confusion matrix – Accuracy, Precision, Recall,
Specificity
F1 score Case study

KNN for regression

Advantages & Disadvantages of KNN
Case study
Optimizing the value of k

Pros and cons of Naïve Baye’s theorem

Case study
 SVM (Support Vector Machines)
Hyperplanes Margin and kernel Cover
Hard vs Soft margin Strength and weakness o
Hinge loss Case study
Kernel tricks

MODULE 4
ENSEMBLE TECHNIQUES
Decision trees
Decision trees for classification and
regression
Entropy
Information gain – Gini Index
Ensemble methods
Introduction to Ensemble methods
Bagging
Machines)
Margin and kernel Cover’s theorem
Strength and weakness of SVM
Case study

Overfitting decision trees Regularization

methods
– Pruning Regression trees Case study
Boosting techniques Ada boosting
Case study
 MODULE 5
UNSUPERVISED LEARNING
K Means clustering

Clustering Importance of scaling

Distance More distance
measures – measures – Mahalanobis
Manhattan, Strength & Weakness of
Euclidean, Visual analysis Silhouette
Chebyshev, clustering Case study
Minkowski
Types of
clustering –
Connectivity
based & Centroid
based
Distance
calculations
Cluster definition
Hierarchical clustering
Lloyd’s algorithm
Connectivity based clustering
Elbow method
Principal component
analysis
Dimensionality reduction
Scaling & Covariance matrix
Eigen value & Eigen vectors

Random Forest classifier

Case study
RNING

Importance of scaling
More distance
measures – Mahalanobis, Jaccard distance
Strength & Weakness of K Means clustering
Visual analysis Silhouette score Dynamic
clustering Case study

Dendrogram Types of Linkages

Elbow method
Feature elimination vs Feature extraction
Case study
MODULE 6
FEATURISATION, MODEL SELECTION & HYPER PARAMET
Feature engineering
Analytics based table Outlier treatment
Model selection and tuning

Variance vs Bias SMOTE

trade-off T Link
Up sampling and
down sampling
Model performance measures
Confusion matrix –
Accuracy, specificity, recall
& precision
ROC curve and AUC curve
Regularising Linear
models
L1 and L2 normalization
Ridge and Lasso regression

ML pipeline
Hands-on on ML pipeline
Bootstrap sampling Grid search CV Randomized
 MODULE 7
RECOMMENDATION SYSTEMS
Introduction to Recommendation systems and its applica
Types of recommendation systems Popularity based reco
system
Market Basket analysis
Concept of Market Basket analysis Support, Confidence and Lift Case
Precision, Recall & Accuracy
Content based recommendation system
Cold start problem
Bag of Words approach TF - IDF

K fold cross validation Implementing K fold cross validation Configuring K

LOOCV
Hands-on – K fold, LOOCV
TION & HYPER PARAMETER

SMOTE
T Link

nce measures
F1 score
Power of the test
Hands-on – Confusion matrix
Hands-on – Ridge and Lasso

eline
ing Grid search CV Randomized search CV
N SYSTEMS
ndation systems and its applications
n systems Popularity based recommendation

ysis Support, Confidence and Lift Case study

ndation system
F

old cross validation Configuring K

 Collaborative filtering
User – User collaborative
filtering
User – Item collaborative
filtering
Matrix factorization based
approach using SVD

Hybrid recommendation systems

MODULE 8
NEURAL NETWORKS BASICS
Introduction to Recommendation systems and its applica

Activation functions Sigmoid function ReLU Softmax Leaky ReLU

Tanh
Gradient descent

Using Gradient descent to minimize loss function Va

batch gradient descent
Mini batching
 Advantages and disadvantages of
collaborative filtering
Case study ROC curve
Precision – Recall curve

ation systems

RKS BASICS
mendation systems and its applications

Softmax Leaky ReLU

dient descent to minimize loss function Vanilla, Stochastic and

ent descent
ng
Learning rate and tuning Sadle points, Hessians a
Learning rate decay
Momentum
Optimizer functions

SGD optimizer Adam optimizer

Introduction to Tensor flow

Understanding
computational graph
Introduction to Keras
Using Keras to build a simple Neural network
Back propagation and chain rule
Jacobian matrix Propagation of loss gradient
Fully connected layer Softmax
Cross – Entropy
Weight initialization
Problem with zero weights initialization Xavier initialization
He initialization
Regularization

Batch Normalization Drop outs

Mini batching Inverted drop outs
 Sadle points, Hessians and local furrows

Adam optimizer
or flow

Using Tensor flow to build a linear regression

model
s
mple Neural network
d chain rule
Propagation of loss gradient
r Softmax

s initialization Xavier initialization

Drop outs
Inverted drop outs
 MODULE 9
COMPUTER VISION
Introduction to Image data
What is a Pixel?
Saving & loading an image file

Introduction to Convolutional Neural Networks

ANN vs CNN
Translational Invariance Convolution operation 2D convolution
Convolution on images
Demonstration of forward and backpropagation in CNNs
Dilated convolution Padding
Pooling
Max pooling and average pooling
Forward and backpropagation in pooling layers Hands-on demo of basic CNN

Babysitting the neural network

Hands-on demo for tuning the hyper parameter for Neural
networks
 Histogram Filters
orks
onvolution

NNs

s-on demo of basic CNN

parameter for Neural

Famous CNN VGGNet - Architecture Ince
architectures Architecture ResNet - Archi
LeNet
ImageNet challenge AlexNet -
Architecture
Transfer Learning

Case study – Transfer Learning on MNIST

Object detection

Object detection Object

localization
RCNN, Fast RCNN and Faster
RCNN
YOLO – Network architecture
Semantic segmentation
What is Semantic segmentation
Recap of CNN Deconvolution in
detail
Other variants of
convolution
1 * 1 convolution Concept of
inception and
separable convolution used
in InceptionNet

Metric Learning
Concept and properties of Using CNNs as a feature e
metric learning
 VGGNet - Architecture InceptionNet -
Architecture ResNet - Architecture

r Learning on MNIST

SSD (Single shot detection) – Network

architecture
Case study – Bounding box regressor
UNet architecture
Case study – One shot learning
Concept of layer wise convolution used in
MobileNet

Using CNNs as a feature extractor

 MODULE 10
NLP

Introduction to NLP Types of problems in NLP Challen

Workflow of dealing with NLP problems Common terminolo
domain
Preprocessing text data

Tokenization
Stop Words removal
Stemming
Bag of Words
Model TF-IDF

Siamese Networks
Siamese Network architecture Distance
measures

Triplet Loss
Why triplet loss is required
Loss function for Siamese networks
Using pre-trained models to build Siamese style networks
Discussion on joint layers
Case study – Arcadian Languages dataset
o NLP Types of problems in NLP Challenges in NLP
ing with NLP problems Common terminology used in NLP

g text data

Lemmatization
Word to features

Distance
measures

ks
Siamese style networks

s dataset
Introduction to Language models
N-grams
Markovian assumption
Word2Vec
Need for deep learning models
CBOW (Continuous Bag of words)
Skip gram
GLOVE
GLOVE as an alternative to
Word2Vec
GLOVE explanation and cost
function

Case study - POS Tagging & Named Entity Recognitio

Introduction to Sequential models
What is sequential data?
Challenges of using traditional ML models to solve sequential dat

Need for memory in neural networks

Types of sequential models – One to many, many to o
many
Recurrent Neural networks (RNNs)
RNN architecture
Backpropagation through time (BPTT) Vanishing & Exploding gra

Long Short Term Memory (LSTM)

ge models
Bi-grams and tri-grams
Comparing the performance of GLOVE and
Word2Vec
Hands-on demo – Word embedding

ing & Named Entity Recognition

tial models
nal ML models to solve sequential data problems

ural networks
odels – One to many, many to one, many to

orks (RNNs)
me (BPTT) Vanishing & Exploding gradients in RNNs

ory (LSTM)
 Concept of gates Forward propagation a
Forget gate backpropagation of LS
Input/ output gate Hands-on – POS taggi
Cell state LSTM

GRU
Applications of LSTMs
Case study – Sentiment analysis using LSTM
Time series analysis
Case study
Case study - Neural Machine Translation Advanced L
architectures
Multi-layer LSTM architecture
Bi-directional LSTM architecture LSTM with attention mechanism

MODULE 11
VISUALIZATION USING TENSORBOARD

What is Tensor board?

Visualizing graphs,
weights, bias & gradients
T-SNE
 Forward propagation and
backpropagation of LSTM
Hands-on – POS tagging using
LSTM

Ms
ment analysis using LSTM
s

Machine Translation Advanced LSTM

cture
hitecture LSTM with attention mechanism

SING TENSORBOARD

Occlusion Experiment Saliency and

Activation maps
Neural Style transfer
 MODULE 12
REINFORCEMENT LEARNING
What is reinforcement learning?
Reinforcement learning framework

Rewards Policy
Goals & Model
actions State

Value-based methods - Q-learning

Q learning – algorithm Q Case study – Smart taxi
table Case study – Frozen lake
Exploration vs
Exploitation SARSA
Q Learning vs SARSA

MODULE 13
INTRODUCTION TO GANS
Introduction to GANs Auto encoders
Deep Convolutional GANs
How to train and common challenges in
GANs

Sample List Of Projects

Identify potential customers for a personal loan product for a bank, allowin
targeted marketing campaigns to increase conversion.
Identify potential customers for a personal loan product for a bank, allowin
targeted marketing campaigns to increase conversion.
 Parkinson’s Disease (PD) is a degenerative neurological disorder marked
dopamine levels in the brain. Traditional diagnosis of onset of Parkinson’s
difficult, as monitoring the onset requires repeated clinic visits by the patie
the project is to build a Machine Learning algorithm over a recording datas
onset of the disease.

The purpose of the case study is to build a model which can classify a give
of three different types of vehicle, using a set of features extracted from th

Build a network intrusion detection system and improve the accuracy of yo

a series of unsupervised and supervised methods.
A Campaign to sell Personal Loans
Bank Note Analysis
Predict the onset of Parkinson’s disease
Silhouette analysis
Network Intrusion detection
Build a recommendation model that can determine and predict the sentime
from text reviews on Amazon’s website.
Recognizing multi-digit numbers in photographs captured at street level is
component of modern-day map making. More broadly, recognizing numbe
a problem of interest to the optical character recognition community. The o
project is to build a python code for image classification from scratch to un
nuances of building and training a model and further to understand the ad
networks.

Facial recognition is a biometric solution that measures unique characteris

face. Given an image or a video capture of a scene with one or more face
designed to detect and classify each face as one of the persons whose ide
known or as an unknown face.
 Fake news is increasingly becoming a threat to our society. It is typically g
increasing commercial interests—attract viewers and collect more advertis
project, you will build a classifier model which can predict whether a piece
using sequential models in Natural Language Processing.

Fake news detection

Face Recognition
Street House View Numbers
Customer Sentiment from Amazon Reviews
CONTACT US+91-8448480528 [email protected] greatlearning.in
EARNING
ning?
mework

Policy
Model

ods - Q-learning
Case study – Smart taxi
Case study – Frozen lake

Semi-supervised GANs
Practical Application of GANs

Of Projects
sonal loan product for a bank, allowing the bank to design
rease conversion.
sonal loan product for a bank, allowing the bank to design
rease conversion.
erative neurological disorder marked by decreased
nal diagnosis of onset of Parkinson’s disease is often
ires repeated clinic visits by the patient. The objective of
ning algorithm over a recording dataset to predict the

uild a model which can classify a given silhouette as one

ng a set of features extracted from the silhouette.

ystem and improve the accuracy of your prediction using

sed methods.

se

can determine and predict the sentiments of the customer

te.
hotographs captured at street level is an important
ng. More broadly, recognizing numbers in photographs is
aracter recognition community. The objective of this
mage classification from scratch to understand the
odel and further to understand the advantages of neural

ion that measures unique characteristics about one’s

ure of a scene with one or more faces, the project is
face as one of the persons whose identity is already
a threat to our society. It is typically generated for
act viewers and collect more advertising revenue. In this
el which can predict whether a piece of news is fake by
anguage Processing.

iews
[email protected] greatlearning.in