-
-
Notifications
You must be signed in to change notification settings - Fork 6
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
ed-dash comments #26
Comments
I've found this lesson to work well with just two features, but I do play around with some of the parameters to demonstrate what is happening. These should be captured in the materials, so I'll try to make some updates to explain things more clearly. |
What I mean is that if we're fitting a random forest to two variables, then I'd expect the feature subsampling to produce trees with one feature, otherwise it's just a regular tree ensemble |
One of the nice things about dealing with only two variables is that we can demonstrate that this expectation is not true for random forests (at least for this particular implementation). If it was true that setting The explanation is that features are being limited at each split, not at the model level: |
Ah. In that case it'd be good to explain that in the lesson |
@alanocallaghan points out that the max_features is confusing for Random Forests. Why does a Random Forest with max_features=1 still result in sub-trees that make decisions based on >1 feature? The explanation is that the max_feature argument is applied at the split level, not the tree level.
@alanocallaghan Please could you take a look at #27 and let me know if this resolves the issue? |
Explain the purpose of max_features for Random Forests. Closes #26.
base_estimator -> estimator in recent sklearn
In the random forest page, we specify
max_features=1
but the decision boundaries are all bivariate. This makes for a very confusing introduction to random forestshttps://carpentries-incubator.github.io/machine-learning-trees-python/06-random-forest/index.html
The text was updated successfully, but these errors were encountered: