Skip to content

fixes when elbo is positive #2092

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Apr 28, 2017
Merged

fixes when elbo is positive #2092

merged 1 commit into from
Apr 28, 2017

Conversation

ferrine
Copy link
Member

@ferrine ferrine commented Apr 28, 2017

For some distributions we do not add normalizing constant (eg Beta) that can result in positive ELBO and wrong inference results. Thus for ELBO based inference methods I just minimize the objective with identity objective transformation.
Transformation does not need to be always identity, there exist methods when square is reasonable choice (LS operator)

@ferrine ferrine merged commit 3a62aec into pymc-devs:master Apr 28, 2017
@ferrine ferrine deleted the fix_positive_elbo branch April 30, 2017 13:17
@fonnesbeck
Copy link
Member

Our beta distribution is normalized (note the subtracted betaln in the logp). There were a few un-normalized logp methods, but I think we fixed them.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants