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.

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

Fixed norm const for SBVM #3411

Merged
merged 2 commits into from
Dec 2, 2024
Merged

Fixed norm const for SBVM #3411

merged 2 commits into from
Dec 2, 2024

Conversation

OlaRonning
Copy link
Member

Ports 1899 from NumPyro. We can use the original solution because torch tensors are not immutable.

The SBVM was not correctly normalized for high concentration because the modified Bessel function was not approximated with enough terms.

New

  • Use terms=max(501, conc1 + conc2) to decide the number of terms to approximate the Bessel function (i.e. log_I1). This heuristic is checked for concentrations in [1,10_000] and not checked formally.
  • Corrected the weighted correlation interval to [-1,1] for SBVM and added a prior suggestion.

@fehiepsi fehiepsi merged commit b34ba6c into dev Dec 2, 2024
9 checks passed
@OlaRonning OlaRonning deleted the fix/sbvm branch December 2, 2024 13:18
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