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

A potential problem with projection on EF members with non-constant base measure #18

Closed

bvdmitri opened this issue Jul 3, 2024 · 1 comment

Assignees

Member

bvdmitri commented Jul 3, 2024

Opening this for historical reasons and so we can close the issue when fixed

From slack @ismailsenoz @Nimrais

The derivation assumes constant base measure. For distributions that do not have constant base measure the expressions used are not correct as the entropy contribution is not as has been derived. There should be gradient of log base measure terms. In the case of Chisq and Poisson because we are projecting from Chi to Chi they cancel out and a happy coincidence gives us the correct result

We should subtract the gradient of log base measure and it should work

related to ReactiveBayes/ExponentialFamily.jl#199

bvdmitri assigned Nimrais

Nimrais mentioned this issue

Logbasemeasure correction #19

Merged

Member Author

bvdmitri commented Jul 4, 2024

Fixed in #19

bvdmitri closed this as completed

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment