Reasoning with Bayesian Networks

Important: Next class Tue 14 Oct, 10-12 (makeup for 17 Sept), at NICTA Seminar Room, Ground Floor, Tower A, 7 London Circuit. Regular time and location from Wed 15 Oct.

This page contains information on the second half of COMP8620 Advanced Topics in AI, offered in Semester 2, 2008 (Phil Kilby discusses search in the first half).

In this part of the course we discuss probabilistic reasoning with Bayesian networks, with a focus on how problem structure can be exploited in various ways to improve the efficiency and scalability of reasoning.

Textbook: Modeling and Reasoning with Bayesian Networks, Adnan Darwiche.

Instructor: Jinbo Huang
Time: Wed 10–12
Place: Graduate Teaching Room, R221, Ian Ross Building

Tentative Schedule

  • [Slides] Week 1: Probability calculus, Bayesian networks
  • [Slides] Week 2: Building Bayesian networks, inference by variable elimination
  • Week 3: Inference by factor elimination, inference by conditioning
  • Week 4: Models for graph decomposition, most likely instantiations
  • Week 5: Complexity of probabilistic inference, compiling Bayesian networks
  • Week 6: Inference with local structure, selected applications
    • Assignment 1, due Wed 15 Oct at beginning of lecture. Complete the following 8 exercises from the textbook: 3.21, 3.22, 4.11, 4.25, 5.1, 5.11, 6.1, 6.4. Late assignments will be penalized by 10% per day, ignoring weekends. Not accepted after beginning of lecture Wed 22 Oct.