documentation of attributes of HistGradientBoostingClassifier except … #16283

judithabk6 · 2020-01-29T14:50:15Z

contributes to #14312
adds documentation for attributes of HistGradientBoostingClassifier, except classes_ (see #16277)

@NicolasHug attribute bin_mapper_ is of private class _BinMapper but is exposed, is that ok?

…classes_ (see scikit-learn#16277)

NicolasHug · 2020-01-29T14:56:03Z

Thanks for the PR @judithabk6

However my impression is that all of these attributes should instead be private rather than public. WDYT @ogrisel?

glemaitre · 2020-01-31T15:58:11Z

sklearn/ensemble/_hist_gradient_boosting/gradient_boosting.py

@@ -923,13 +923,24 @@ class HistGradientBoostingClassifier(BaseHistGradientBoosting,

    Attributes
    ----------
+    bin_mapper_ : BinMapper object


IMO, this could be private.

glemaitre · 2020-01-31T15:58:41Z

sklearn/ensemble/_hist_gradient_boosting/gradient_boosting.py

@@ -923,13 +923,24 @@ class HistGradientBoostingClassifier(BaseHistGradientBoosting,

    Attributes
    ----------
+    bin_mapper_ : BinMapper object
+        Transformer that maps a dataset into integer-valued bins.
+    do_early_stopping_ : bool


This could be public since it is derived from an input parameter.

Not very useful, also obsolete once #14516 is merged

agreed this is obsolete now :)

glemaitre · 2020-01-31T15:58:48Z

sklearn/ensemble/_hist_gradient_boosting/gradient_boosting.py

+        Transformer that maps a dataset into integer-valued bins.
+    do_early_stopping_ : bool
+        Specifies if early stopping is used during the training.
+    loss_ : callable


This could be public since it is derived from an input parameter.

I disagree, the losses are all private objects

The effective loss name could be stored as a public attribute.

glemaitre · 2020-01-31T15:58:56Z

sklearn/ensemble/_hist_gradient_boosting/gradient_boosting.py

+        Specifies if early stopping is used during the training.
+    loss_ : callable
+        Loss function used by the algorithm.
+    n_features : int


This could be private

Private for now and replaced by n_features_in_ later.

glemaitre · 2020-01-31T15:59:16Z

sklearn/ensemble/_hist_gradient_boosting/gradient_boosting.py

    n_iter_ : int
        The number of estimators as selected by early stopping (if
        n_iter_no_change is not None). Otherwise it corresponds to max_iter.
    n_trees_per_iteration_ : int
        The number of tree that are built at each iteration. This is equal to 1
        for binary classification, and to ``n_classes`` for multiclass
        classification.
+    scorer_ : callable


This could be public

I don't think this is a very useful public attribute. What are the benefit of exposing it?

Indeed, we could also keep it private. The only other classes that have a public scorer_ attribute are the SearchCV classes and for those it's actually a list of scorer objects to deal with multi-metrics things.

Let's keep it private.

ogrisel

Indeed we can probably close this PR and instead make those attributes private by moving the underscore at the start in the name of the attributes.

ogrisel · 2020-02-13T09:56:40Z

sklearn/ensemble/_hist_gradient_boosting/gradient_boosting.py

+        Transformer that maps a dataset into integer-valued bins.
+    do_early_stopping_ : bool
+        Specifies if early stopping is used during the training.
+    loss_ : callable


The effective loss name could be stored as a public attribute.

ogrisel · 2020-02-13T09:59:15Z

sklearn/ensemble/_hist_gradient_boosting/gradient_boosting.py

+        Specifies if early stopping is used during the training.
+    loss_ : callable
+        Loss function used by the algorithm.
+    n_features : int


Private for now and replaced by n_features_in_ later.

ogrisel · 2020-02-13T10:02:29Z

sklearn/ensemble/_hist_gradient_boosting/gradient_boosting.py

    n_iter_ : int
        The number of estimators as selected by early stopping (if
        n_iter_no_change is not None). Otherwise it corresponds to max_iter.
    n_trees_per_iteration_ : int
        The number of tree that are built at each iteration. This is equal to 1
        for binary classification, and to ``n_classes`` for multiclass
        classification.
+    scorer_ : callable


Indeed, we could also keep it private. The only other classes that have a public scorer_ attribute are the SearchCV classes and for those it's actually a list of scorer objects to deal with multi-metrics things.

Let's keep it private.

cmarmo · 2020-03-28T14:33:43Z

@ogrisel, @glemaitre, @NicolasHug , is a consensus there about making all the attributes in this PR private? If so, @judithabk6 are you ok in proposing a new PR in that direction? Feel free to ask for clarifications if you need some. Thanks for your patience!

NicolasHug · 2020-03-28T14:38:12Z

Yes let's make everything private except for do_early_stopping_ since it's useful now that we have early_stopping='auto' by default

documentation of attributes of HistGradientBoostingClassifier except …

25ec237

…classes_ (see scikit-learn#16277)

jeremiedbb added the Sprint label Jan 29, 2020

glemaitre reviewed Jan 31, 2020

View reviewed changes

ogrisel reviewed Feb 13, 2020

View reviewed changes

cmarmo added the Needs Decision Requires decision label Feb 16, 2020

github-actions bot added the module:ensemble label Mar 2, 2020

cmarmo removed the Needs Decision Requires decision label Mar 28, 2020

cmarmo added Easy Well-defined and straightforward way to resolve good first issue Easy with clear instructions to resolve help wanted Stalled labels Jun 12, 2020

cmarmo removed the help wanted label Jun 23, 2020

This was referenced Jun 23, 2020

Documentation of attributes of HistGradientBoostingClassifier #17678

Merged

DOC add missing attributes in HistGradientBoostingRegressor #17677

Merged

glemaitre closed this in #17678 Jun 24, 2020

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

documentation of attributes of HistGradientBoostingClassifier except … #16283

documentation of attributes of HistGradientBoostingClassifier except … #16283

judithabk6 commented Jan 29, 2020

NicolasHug commented Jan 29, 2020

glemaitre Jan 31, 2020

glemaitre Jan 31, 2020

NicolasHug Jan 31, 2020

glemaitre Feb 12, 2020

glemaitre Jan 31, 2020

NicolasHug Jan 31, 2020

ogrisel Feb 13, 2020

glemaitre Jan 31, 2020

ogrisel Feb 13, 2020

glemaitre Jan 31, 2020

NicolasHug Jan 31, 2020

ogrisel Feb 13, 2020

ogrisel left a comment

ogrisel Feb 13, 2020

ogrisel Feb 13, 2020

ogrisel Feb 13, 2020

cmarmo commented Mar 28, 2020

NicolasHug commented Mar 28, 2020

documentation of attributes of HistGradientBoostingClassifier except … #16283

documentation of attributes of HistGradientBoostingClassifier except … #16283

Conversation

judithabk6 commented Jan 29, 2020

NicolasHug commented Jan 29, 2020

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ogrisel left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

cmarmo commented Mar 28, 2020

NicolasHug commented Mar 28, 2020