ML4001 Probabilistic Graphical Models Syllabus:

ML4001 Probabilistic Graphical Models Syllabus – Anna University PG Syllabus Regulation 2021

COURSE OBJECTIVES:

 To understand basic concepts of probabilistic graphical models
 To explore different aspects of representation of probabilistic graphical models
 To study different inference techniques
 To apply various inference techniques
 To understand learning associated with probabilistic graphical models

UNIT I INTRODUCTION

Probabilistic Graphical Models – Motivation –Foundations – Probability Theory –Graphs – Independence Properties – Bayesian Network Representation – Independence in Graphs – From Distribution to Graphs

UNIT II REPRESENTATION

Undirected Graphical Models – Parameterization –Markov Network Independencies – Bayesian Networks and Markov Networks – Local Probabilistic Models – Tabular CPDs – Template –Based Representation – Temporal Models- Exponential Family – Entropy and Relative Entropy

UNIT III INFERENCE

Exact Inference – Variable Elimination- Conditioning – Clique Trees – Message Passing – Inference as Optimization – Exact Inference as Optimization – Propagation based Approximation

UNIT IV ADVANCED INFERENCE

Particle Based Approximate Inference – Forward Sampling – Markov Chain Monte Carlo Methods – Map Inference – Variable Elimination for Map – Max-Product in Clique Trees – Exact Inference in Temporal Models

UNIT V LEARNING

Learning Graphical Models – Overview – Goals – Learning Tasks –Maximum Likelihood Estimation for Bayesian Networks – Bayesian Parameter Estimation – Structure Learning in Bayesian Networks -Methods –Learning Undirected Models

SUGGESTED ACTIVITIES:

1. Problems in Probability
2. Design examples of Probabilistic Graphical Models
3. Hand simulate all inferences possible with graphical models for examples of your choice
4. Give an example for temporal probabilistic graphical model
5. Discuss pros and cons of different learning techniques

COURSE OUTCOMES:

CO1: Understand basic concepts of probabilistic graphical models
CO2: Automatically convert a problem into a probabilistic graphical model
CO3: Implement a simple graphical model
CO4: Understand issues associated with temporal models
CO5: Design a learning system for the graphical model

TOTAL:45 PERIODS

REFERENCES

1. D. Koller and N. Friedman, “Probabilistic Graphical Models: Principles and Techniques”, MIT Press, 2009.
2. Probabilistic Machine Learning: An Introduction by Kevin Patrick Murphy.MIT Press, March 2022.
3. M.I. Jordan, “An Introduction to Probabilistic Graphical Models”, Preprint.
4. C.M. Bishop, “Pattern Recognition and Machine Learning”, Springer, 2006.
5. K.P. Murphy, “Machine Learning: A Probabilistic Perspective”, MIT Press, 2012.
6. David Barber. “Bayesian Reasoning and Machine Learning”, Cambridge University Press. 2012.
7. David Mackay, “Information Theory, Inference, and Learning Algorithms”, Cambridge university press. 15 February 2010.