Add test for MAP/MLE estimation #2484
Conversation
- float32 support - different arg, method="L-BFGS-B" instead of fmin=optimize.fmin_l_bfgs_b - allow keyboard interrupt, save current value - returns untransformed and transformed variables - removes monitor, adds tqdm progress bar
| vars : list | ||
| List of variables to set to MAP point (Defaults to all continuous). | ||
| fmin : function | ||
| Optimization algorithm (Defaults to `scipy.optimize.fmin_bfgs` unless |
There was a problem hiding this comment.
Do we need to deprecate fmin rather than getting rid of it completely right away? This might cause a fair amount of breakage. Thoughts?
There was a problem hiding this comment.
I think deprecate it is a better idea. @bwengals
There was a problem hiding this comment.
Maybe just look for it in kwargs and warn if its there.
There was a problem hiding this comment.
I'm happy to re-work this PR either choice on the breakage, but there wasn't really a decision made in #2468...
There was a problem hiding this comment.
maybe it's easier after #2468 is merged.
There was a problem hiding this comment.
Checking for fmin in kwargs would work I think. Should it warn and then run?
There was a problem hiding this comment.
actually maybe its a bit trickier than I thought. different fmin functions have slightly different APIs. According to its docstring, fmin_bfgs doesnt support a cost function that returns both the value and gradient, while fmin_l_bfgs_b does.
|
So we don't have to comment on the same API changes twice. |
4 participants
find_MAP().