progress.txt

Main progress:
 - Designed a simple DSL which can be used to describe generative models
 - Distributions are modeled as a monad to allow composition of Distributions
 - DSL is shallowly embedded, so distributions can be composed using OCaml language constructs including:
    - functions/lambdas
    - recursion
    - if/then/else
 - Currently a set of primitive distributions - categorical, uniform, binomial, etc. 
 - Inference algorithms have been implemented:
    - exact inference (bfs)
    - importance sampling
    - rejection sampling
    - metropolis-hastings
    - sequential monte-carlo (particle filter)
    - particle cascade
    - particle independent metropolis hastings
 - set of example models - e.g. hmm, linear regression, geometric (recursive), coins, sprinkler, etc.