Thomas J. Fan
Thomas J. Fan
Target encoder does not work better: ```py from sklearn.compose import ColumnTransformer from sklearn.pipeline import Pipeline from category_encoders import TargetEncoder from sklearn.preprocessing import StandardScaler from sklearn.linear_model import Ridge from sklearn.model_selection import...
As noted in https://github.com/scikit-learn/scikit-learn/issues/24131#issuecomment-1211756677, this was fixed in #24168
Ah I see, there are two issues. I'm okay with keeping this open.
Yup, let's follow up with a PR on improving the the example. I think the formatting suggested in https://github.com/scikit-learn/scikit-learn/pull/24104#discussion_r966071617 is sufficient.
The dependencies are in `setup.py` through: https://github.com/scikit-learn/scikit-learn/blob/f0cbdc15f5031f0595ffcc8f822e5318a2335349/setup.py#L276 which dynamically imports https://github.com/scikit-learn/scikit-learn/blob/main/sklearn/_min_dependencies.py to get dependencies for install. Currently, `_min_dependencies` is the "single source of truth" for all our dependencies. Strictly speaking,...
Before `pyproject.toml` existed, I think most package managers list out the build and runtime dependencies themselves by looking at the `README`. For example, [ubuntu's APT](https://packages.ubuntu.com/jammy/python3-sklearn) lists out scikit-learn's dependencies themselves...
> No. setup.py had/has tags install_require, test_require, etc. scikit-learn already populates `install_require` in `setup.py` dynamically. When we build a wheel for PyPI, this metadata is injected into the wheel.
> This is a very uncommon behavior in scikit-learn. Most others list them in setup.py and pyproject.toml I do not think there has been a common standard to specify this...
I agree the term "third party" is confusing here. I am in favor with your suggestion of using "domain specific groups"
Note that since `_set_tree_class` and other extension points will be private in scikit-learn. This means that they can change without deprecation between scikit-learn versions. A third party library that subclasses...