sample generative model

[martinjrobins]

each pints log-likelihood defines a generative model, which you can sample (e.g. for simulating fake data), it would be nice if there was a method to sample from each log-likelihood, given a parameter set

[ben18785]

Hmm, interesting. Would it be an issue though that a log-likelihood contains data, so a user might think it's somehow sampling from a sorta posterior distribution? (Whereas I guess it would just be sampling data from a distribution with the parameters supplied by the user?)

[HOLL95]

@martinjrobins would you mind clarifying what you meant by "sample from each log-likelihood" and how it would differ from e.g. problem.evaluate(random_values)?