Fix gmm initialization from weighted samples

jmschrei / pomegranate

Fast, flexible and easy to use probabilistic modelling in Python.

http://pomegranate.readthedocs.org/en/latest/

MIT License

3.35k stars 589 forks source link

Closed gerwang closed 1 year ago

gerwang commented 1 year ago

This fixes a bug that empirically lowers the quality of the initialization.

When initializing gmm from weighted samples, the priors should be proportional to the sum of sample weights.