Senior Data Scientist, @UPennIBI. #MachineLearning & #AI. Data tinkerer. Community leader for @DataIsBeautiful. #OpenScience advocate.

Search Tweets

Twitter Feed

@randal_olson — 112,244 followers, 16,213 tweets

Sep 25, 2017

@amuellerml Currently running [0., 0.05] every 0.0025 to see if setting != 0 is useful. I suppose impurity decreases can be small & still useful.

Sep 25, 2017

I've been trying [0, 0.5] in increments of 0.05 and setting != 0 pretty much always seems to be worse.

Sep 25, 2017

Does anyone have a good sense of the param range to try for the "min_impurity_decrease" param in #sklearn tree-based models?

Show more