A PGM/bayes net library designed with a particular goal in mind:
If x and y are random variables, then z = x + y should be all that's needed to
compute a new random variable that is distributed as their sum, even taking into account
their joint distribution. No special syntax, new operators, or do-blocks should be necessary.
This library accomplishes this task with probabilistic graphical models for all operations
in Num over the Rationals (and thus any composition of those functions), serving as a
proof of concept for a more general extension. An example is in Main.