search

sseemayer / mixem

Pythonic Expectation-Maximization (EM) implementation for fitting mixtures of probability densities
https://mixem.readthedocs.org/en/latest/
26 stars 16 forks source link

[] return full history. Stable logsumexp #3

Closed shouldsee closed 5 years ago

shouldsee commented 5 years ago

Thank @sseemayer for this great framework.

Would like to suggest

  1. using a stable logsumexp().
  2. return full training history from mixem.em

Maybe we should also write some tests.

sseemayer commented 5 years ago

Thanks a lot for the enhancements! I'm not using the library any more but I'm glad that it's useful to you and hope that merging this will help.

shouldsee commented 5 years ago

@sseemayer My pleasure. Did you find any other good Python framework for EM models at all? Or what's the exact reason this library wasn't used?

sseemayer commented 5 years ago

I would probably still use mixem if I wanted to train EM models but am just not currently having that particular need 🙂