Haoyin Xu

To put it more specifically, the original UF has the |_D_P| : |_D_V| : |_D_E| ratio as `0.4 : 0.3 : 0.3`, but the new benchmark functions has `0.4*0.7 :...

I personally think that modifying the existing function makes more sense, so it could save us an additional splitter class?

@rflperry I hold some objections for the copying idea, as copying the classifier code basically means copying over the entire sklearn.tree [folder](https://github.com/scikit-learn/scikit-learn/tree/master/sklearn/tree). I found it very hard to locate the...

@v715 I can confirm that feature importance is absent from the current implementation. @mkusman1 Always feel free to open new issues after confirming no repetition~

> For the sake of code-review culture: > > > minor changes for figure color and code block formats > > What do the changes induce? And why are they...

Jovo wants to add a figure similar to[ this one](https://www.researchgate.net/figure/The-null-distribution-of-the-test-statistic-in-4-for-the-null-hypothesis-that-there-is_fig1_304858060), which would require all the intermediate statistics (the ones that 2 permuted forest make). `build_coleman_forest` doesn't return those values, unfortunately,...

Probably? Nothing has been done about it.

I found some segmentation fault bug with cython on tree level partial_fit, trying to resolve it.