CP7002 PROBABILISTIC REASONING SYSTEMS - ANNA UNIVERSITY 1ST SEM CSE SYLLABUS REG-2013

ANNA UNIVERSITY, CHENNAI

REGULATIONS - 2013

M.E. COMPUTER SCIENCE AND ENGINEERING

CP7002 PROBABILISTIC REASONING SYSTEMS

OBJECTIVES: 
 To construct and reason with Bayesian networks
 To reason with temporal models
 To make exact and approximate inferences with graphical models
 To understand learning of parameters for probabilistic graphical models
 To understand actions and decisions with probabilistic graphical models

UNIT I REPRESENTATION
Probability Theory, Graphs, Bayesian network representation: Bayes networks, Independence in graphs – Undirected graphical models: Parameterization, Markov Network independencies – Conditional Bayesian networks.

UNIT II TEMPLATE BASED REPRESENTATION
Temporal models (Dynamic Bayesian networks , Hidden Markov Models) – Directed probabilistic models for object-relational domains – Inference in temporal models: Kalman filters.

UNIT III INFERENCE
Exact inference: Variable elimination – Exact inference: Clique trees (Junction trees) – Approximate inference: Forward sampling, Importance sampling, MCMC – MAP inference:
Variable elimination for MAP, Max-product in clique trees.

UNIT IV LEARNING
Learning graphical models – Parameter estimation: maximum-likelihood estimation, MLE for
Bayesian networks, Bayesian parameter estimation – Structure learning in Bayesian networks:
Constraint based, structure scores, structure search – Partially observed data: Parameter
estimation, Learning models with hidden variables – Learning undirected models: Maximum
likelihood

UNIT V ACTIONS AND DECISIONS
Causality – Utilities and decisions – Structured decision problems

TOTAL: 45 PERIODS

OUTCOMES:
Upon Completion of the course, the students will be able to
 Construct Bayesian networks
 Reason with Bayesian networks
 Reason with Dynamic networks and Hidden Markov Models
 Conduct inferences with Bayesian networks
 Implement algorithms to learn probabilistic graphical models
 Explain actions and decisions with probabilistic graphical models

REFERENCES:
1. Daphne Koller and Nir Friedman, “Probabilistic Graphical Models: Principles and Techniques”,MIT Press, 2009.
2. David Barber, “Bayesian Reasoning and Machine Learning”, Cambridge University Press,
2012.27
3. Adnan Darwiche, “Modeling and Reasoning with Bayesian networks”, Cambridge University
Press, 2009.
4. Kevin P. Murphy, “Machine Learning: A Probabilistic Perspective”, MIT Press, 2012.
5. Stuart Russel and Peter Norvig, “Artificial Intelligence: A Modern Approach”, Third Edition,
Prentice Hall, 2009.