Chapter 10. Factored inference algorithms

 

This chapter covers

  • The basics of factored inference, definition of factors, and operations on factors
  • The variable elimination algorithm
  • The belief propagation algorithm

Now that you understand the basic rules of probabilistic inference, you’ll spend the next two chapters learning about some of the inference algorithms used in probabilistic programming. This will give you better insight into which algorithm works best for a particular problem and how to design a model sympathetic to that algorithm.

There are two main types of inference algorithms:

  • Factored algorithms work by operating on data structures called factors that capture the probabilistic model being reasoned about.
  • Sampling algorithms work by creating examples of possible worlds from the probability distribution and using those examples to answer queries.

You’ll look at sampling algorithms in the next chapter. This chapter covers factored algorithms by looking at the following:

10.1. Factors

10.2. The variable elimination algorithm

10.3. Using VE

10.4. Belief propagation

10.5. Using BP

10.6. Summary

10.7. Exercises