FIX Raise error when missing-values encountered in scikit-tree trees #264
Conversation
Signed-off-by: Adam Li <adam2392@gmail.com>
Signed-off-by: Adam Li <adam2392@gmail.com>
Signed-off-by: Adam Li <adam2392@gmail.com>
| @@ -64,3 +64,7 @@ def compute_similarity_matrix(self, X): | |||
| The similarity matrix among the samples. | |||
| """ | |||
| return compute_forest_similarity_matrix(self, X) | |||
|
|
|||
| def _more_tags(self): | |||
| # XXX: no scikit-tree estimators support NaNs as of now | |||
There was a problem hiding this comment.
Scikit-Tree appears to be based on Scikit-Learn 1.3.X, which supports missing values in standalone decision tree models (DecisionTreeClassifier, DecisionTreeRegressor) by default. The support for missing values in random forest models (RandomForestClassifier and RandomForestRegressor) was added in 1.4.X.
In that regard, perhaps the _more_tags method should be base class-dependent (one tag for DecisionTree and another one for RandomForest)?
There was a problem hiding this comment.
Scikit-tree will depend on each version of scikit-learn successively very tightly for now, so we'll consider the next release of v0.8 to be tied to 1.4.x. Is that okay?
There was a problem hiding this comment.
Right now, when there is no missing value support, then it would be okay to attach the same "disable all"-type_more_tags method to the tree ABC, so that it affects all tree and tree ensemble estimators.
The exact SkTree <-> SkLearn version matching is not critical until SkTree really starts enabling missing values.
|
I will merge this if LGTYou @vruusmann? |
Signed-off-by: Adam Li <adam2392@gmail.com>
|
CI failures seem unrelated: actions/setup-python#855 |
Fixes #263
Changes proposed in this pull request:
Before submitting
section of the
CONTRIBUTINGdocs.Writing docstrings section of the
CONTRIBUTINGdocs.After submitting