Normalize posteriors returned by HMMClassifier
eonu opened this issue · 1 comments
eonu commented
Currently HMMClassifier
returns unnormalized posteriors for each class when return_scores=True
in predict()
.
It isn't much work to normalize these posteriors by simply dividing by the sum of all class posteriors, and having a true class distribution as an output is useful.
eonu commented
This cannot be done easily due to how small the log posteriors are.