#Graphical Adaptive Sampler

This project is (going to be) a combination of discrete belief propagation and adaptive rejection sampling.

I plan to implement belief propagation on tree Factor Graphs for discrete distributions. And then using adaptive rejection sampling apply this to simple (univariate) continuous distributions. The idea is loosely based off of a Hackathon project undertaken whilst at MSR.

This is a work in progress.

• Adaptive Rejection Sampling

• Belief Propagation

1. Factor Graphs and the Sum-Product Algorithm
2. The OS* Algorithm: a Join approach to Exact Optimization and Sampling