search

david-cortes / contextualbandits

Python implementations of contextual bandits algorithms
http://contextual-bandits.readthedocs.io
BSD 2-Clause "Simplified" License
750 stars 148 forks source link

warm_start for online policy #66

Closed jordantkohn closed 1 year ago

jordantkohn commented 1 year ago

I don't see how the warm_start parameter in _BasePolicy changes the behavior of fit().

When warm_start = True in _BasePolicy.fit(), use_warm can become True. use_warm is then passed to _OneVsRest's instantiation as parameter warm, where it does not appear inside the class definition.

Am I missing something?

david-cortes commented 1 year ago

Yes, thanks for the bug report. This was a vestige of older code that I've forgotten to remove. It currently doesn't make any difference as passing warm_start=True to the underlying base estimator would suffice. PR to remove it would be welcome.