google-deepmind / google-deepmind/PGMax

MAX factor

Open
#9 0 comments 0 reactions 0 assignees View on GitHub
Dominant language
Jupyter Notebook
Stars
166
Forks
12
PR merge metrics
No merged PRs in 30d

Description

Would it be possible to implement a MAX factor in PGMax? The OR factor works well for binary variables, but in some cases it makes sense to use more than two states for a variable, and the OR factor isn't suitable in such cases. MAX would be equivalent to OR for binary variables, but it would also work for non-binary ones.

My use case: We have a Bayesian network that mostly consists of Noisy-MAX nodes. Most of them are binary, and those can be implemented with an OR factor (with some intermediate variables for the noisiness), but some of them are not binary. We can use enum factors to implement these nodes, but I suspect it would be more efficient to use MAX factors.

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.