parrt
Add random forest 2D feature space visualization
I added two interesting methods that show the 2D feature space for classifiers associated with one or more trees. One shows the regions fused together, ctreeviz_bivar_fusion(), and the other does a nice tiling to show probabilities or predicted values for the grid, crfviz_bivar().
Heh, @tlapusan can you take a look at this and the notebook? This is my first draft at the API but I think it's okay. Would be very happy to hear about recommendations for the interface or its functionality. It has a number of limitations at this point, such as being limited to sklearn or any other model that can answer method predict_proba(). Also assumes continuous class target values 0..k-1.
I'm going to leave this alone for a while until I can finish some other projects; let's focus on getting your PR merged and released... I think I'm still behind on your LightGBM PR. going to look for that one now
I had a commit that had the rename for _normalize ... just pushed that.
Also I realize that it's actually Pillow not PIL. in fact the documentation for pillow says we need to remove PIL before installing pillow.
@tlapusan I think we can close this one right? I think we have the two dimensional future space visualization in.
to be honest, I don't remember too much but I see that the code was merged into master. So we should be ok to close it.