SPEAKER: Andrzej Wąsowski, Professor at ITU.
ABSTRACT: I will present the main design decisions and some examples of a simple
probabilistic programming library, Probula, that I have developed to
teach Advanced Programming. The main ingredients are:
– constructors for main discrete and continuous distributions in an
internal DSL style, relying heavily on extension methods and
overloading combined with a monadic encoding
– an (almost) pure encoding of sample generation
– an importance sampling inference algorithm
– output of posteriors as CSV files for plotting
The talk serves as a stepping stone towards developing a lecture for
Advanced Programming on this topic. The lecture takes place ca. 3 weeks
later.