Scribd
Upload
446 views

Machine Learning Paradigms

Machine learning involves algorithms that improve performance on tasks through experience. There are supervised learning algorithms that learn from labeled training data, unsupervised learning algorithms that learn without labels, and reinforcement learning algorithms that learn from rewards and punishments. Some common machine learning algorithms include k-nearest neighbors, decision trees, random forests, support vector machines, neural networks, k-means clustering, and hierarchical clustering. These algorithms are applied to problems like classification, regression, clustering, and reinforcement learning.

Uploaded by

teja ayyagari
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
446 views

Machine Learning Paradigms

Machine learning involves algorithms that improve performance on tasks through experience. There are supervised learning algorithms that learn from labeled training data, unsupervised learning algorithms that learn without labels, and reinforcement learning algorithms that learn from rewards and punishments. Some common machine learning algorithms include k-nearest neighbors, decision trees, random forests, support vector machines, neural networks, k-means clustering, and hierarchical clustering. These algorithms are applied to problems like classification, regression, clustering, and reinforcement learning.

Uploaded by

teja ayyagari
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 27

MACHINE LEARNING

PARADIGMS
WHAT IS MACHINE LEARNING
 “Learning is any process by which a system improves
performance from experience.” -
Herbert Simon

Definition by Tom Mitchell (1998):


Machine Learning is the study of algorithms that
 improve their performance P
 at some task T
 with experience E.

 A well-defined learning task is given by <P, T, E>.


AI VS ML VS DL
Traditional Programming

Data
Computer Output
Program

Machine Learning

Data
Computer Program
Output
APPLICATIONS OF ML ALGORITHMS
Web search Vision processing
Computational biology Language processing
Finance
E-commerce Forecasting things like
Space exploration stock market trends,
Robotics weather
Information extraction Pattern recognition
Social networks
Debugging Games
Data mining
Expert systems
Robotics
LEARNING SYSTEM MODEL

Testing

Input Learning
Samples Method

System

Training
TRAINING AND TESTING

Data acquisition Practical usage

Universal set
(unobserved)

Training set Testing set


(observed) (unobserved)
TRAINING AND TESTING
 Training is the process of making the system able to learn.
 No free lunch rule:
 No Algorithm is best
 No Universal best Algorithm
 Training set and testing set come from the same distribution
 Need to make some assumptions or bias
PERFORMANCE
 There are several factors affecting the performance:
 Types of training provided
 The form and extent of any initial background knowledge
 The type of feedback provided
 The learning algorithms used

 Two important factors:


 Modeling
 Optimization
ALGORITHMS
 The success of machine learning system also depends on
the algorithms. 

 The algorithms control the search to find and build the


knowledge structures.

 The learning algorithms should extract useful


information from training examples.
MACHINE LEARNING ALGORITHMS
MACHINE LEARNING ALGORITHMS
MACHINE LEARNING ALGORITHMS
 Supervised (inductive) learning
 Given: training data + desired outputs (labels)
 Unsupervised learning
 Given: training data (without desired outputs)
 Semi-supervised learning
 Given: training data + a few desired outputs
 Reinforcement learning
 Rewards from sequence of actions
MACHINE LEARNING ALGORITHMS
 Regression : How much and when
 How much stock price
 Mow may tickets will be sold in in 2019

 Classification: Yes or No
 Is fraudulent transaction
 Will this engine fail

 Deep learning:
 Image recognition
 Speech recognition
SUPERVISED LEARNING
SUPERVISED LEARNING
 Given:
 Trainingdata: (x1; y1)… (xn; yn) = xi ∈ Rd and yi is the label.
 example x1 → x11 x12 … x1n = y1 ← label
 Learning a model from labeled data.
SUPERVISED LEARNING
 When the output y is one of a finite set of values (such as
sunny, cloudy or rainy), the learning problem is called
classification
 Boolean or binary classification: if there are only two values.
 When y is a number (such as tomorrow's temperature),
the learning problem is called regression
CLASSIFICATION

 Example: Credit scoring


 Differentiating between low-
risk and high-risk customers
from their income and savings

Discriminant: IF income > θ1 AND savings > θ2


THEN low-risk ELSE high-risk
CLASSIFICATION: APPLICATIONS
 Pattern recognition
 Face recognition: Pose, lighting, occlusion (glasses,
beard), make-up, hair style
 Character recognition: Different handwriting styles.

 Speech recognition: Temporal dependency.

 Use of a dictionary or the syntax of the language.

 Sensor fusion: Combine multiple modalities; eg, visual


(lip image) and acoustic for speech
 Medical diagnosis: From symptoms to illnesses
REGRESSION
 Example: Price of a
used car
 x : car attributes

y : price
y = g (x | θ )
g ( ) model,
θ parameters
UNSUPERVISED LEARNING
UNSUPERVISED LEARNING
 Training data: examples x.
 x1..... Xn, xi ∈ R n
 Learning a model from unlabeled data.
 Clustering/segmentation:
f : Rd → { C1….Ck} (set of clusters).
 Example: Find clusters in the population, fruits, species.
UNSUPERVISED LEARNING
 Learning “what normally happens”
 Clustering: Grouping similar instances

 Example applications
 Customer segmentation in CRM
 Image compression: Color quantization
 Bioinformatics: Learning motifs
REINFORCED LEARNING

 Given a sequence of states and actions with (delayed)


rewards, output a policy
 Policy is a mapping from states to actions that tells you
what to do in a given state
 Examples:
 Game playing
 Robot in a maze
 Balance a pole on your hand
REINFORCEMENT LEARNING
 Learning a policy: A sequence of outputs
 No supervised output but delayed reward

 Credit assignment problem

 Game playing

 Robot in a maze

 Multiple agents, partial observability, ...


MACHINE LEARNING ALGORITHMS…
SUPERVISED LEARNING
 Classification and Regression is done by various
algorithms like…

 K-Nearest Neighbors
 Decision Trees

 Random Forests

 Support Vector Machines

 Neural Networks
MACHINE LEARNING ALGORITHMS…
UNSUPERVISED LEARNING
 K-Means Clustering
 Hierarchical Clustering

 Apriori algorithm for association rule learning problem

You might also like