Skip to content
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.

Sign up for GitHub

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

Jump to bottom

Optimization reproducibility #2

Open
pedrofale opened this issue Aug 9, 2022 · 0 comments
Open

Optimization reproducibility #2

pedrofale opened this issue Aug 9, 2022 · 0 comments

Comments

@pedrofale
Copy link
Member

pedrofale commented Aug 9, 2022 •
edited
Loading

Running SCATrEx twice with the same random seed yields different results. This does not happen when we do not do any ELBO optimization, and just change the parameters in the move proposals and do ELBO evaluations with those. It is irrespective of local in the run_search being True or False. For some reason, some moves lead to ELBO optimization steps with small divergences between runs, and this accumulates leading to different trees in the end.

Commenting out the ancestor nodes' contribution from the node weights seems to fix it. So it might be a numerical instability problem that XLA does not deal with deterministically?
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@pedrofale