search

choderalab / pinot

Probabilistic Inference for NOvel Therapeutics
MIT License
15 stars 2 forks source link

modularize `pinot.Net` to enable non-gradient-based optimization #14

Open yuanqing-wang opened 4 years ago

yuanqing-wang commented 4 years ago

@karalets

Could you

here https://github.com/choderalab/pinot/tree/theo_sandbox

by next Fri?

karalets commented 4 years ago

I vote we do that afterwards and stick to gradient based method (and samplers) for now.

This is going to mostly require changes to the experimental infrastructure, not so much to models, as models do not change based on inference, but you will see that clearer soon.