search

alashworth / test-issue-import

0 stars 0 forks source link

Expose Laplace Approximation #195

Open alashworth opened 5 years ago

alashworth commented 5 years ago

Issue by charlesm93 Friday May 04, 2018 at 16:31 GMT Originally opened as https://github.com/stan-dev/stan/issues/2522

Summary:

Complements issue 755 in the math repo.

Description:

The goal is to design specialized functions to efficiently compute the posteriors of latent gaussian models. In particular we want to consider the case where the observations follow a conditional distribution which is Normal, Poisson, Binomial, and Negative Binomial.

Exposing the functions shouldn't be too hard, as at first glance the functions are not higher-order functions. This may change if we make the functions less specialized and more flexible.

Current Version:

v2.17.1

alashworth commented 5 years ago

Comment by bob-carpenter Monday May 28, 2018 at 22:00 GMT

@charlesm93 --- thanks for following up on this. Could you include the specific signatures you're proposing either here or in the math lib issue?

I'd strongly suggest tackling one of these functions end-to-end first. At that point, the others should be easy to knock out or delegate.

alashworth commented 5 years ago

Comment by charlesm93 Wednesday May 30, 2018 at 13:52 GMT

@bob-carpenter Yes, I'll get to it, though not in the immediate future. We still need a good proof of concept and tests to asses how efficacious the approximation is. But right now, I don't have time to get to it.