Meeting 17: Bayesian Reasoning

Reading: AIAMA 13.5 see also Model Comparison, MacKay Chapter 28 (see link to pdf on website under Text and Resources)

Bayes rule is a very powerful tool for reasoning. It is an optimal procedure for updating beliefs and doing inference. It forms the basis of many algorithms in AI, robotics, and machine learning, and is even a contending theory for how our brains work.

Questions you should be able to answer after reading are:

  1. What is the difference between diagnostic and causal reasoning?
  2. What are evidence variables?
  3. What are hidden variables?
  4. What is the likelihood?
  5. What is the prior?
  6. What is the posterior?
  7. What is the model evidence?
  8. How is Bayes used to update beliefs given evidence?
  9. How is Bayes used to make decisions?
  10. How is Bayes used to compare different agent models?
  11. How does the factoring of the joint indicate conditional independence of variables?
  12. What is a naive Bayes model?