[MRG] Adds Plotting API to Partial Dependence #14646

thomasjpfan · 2019-08-13T20:37:06Z

Reference Issues/PRs

What does this implement/fix? Explain your changes.

Adds PartialDependenceDisplay for plot_partial_dependence.
Deprecates the fig option. The new API can take a single axes and place all the plots in that ax.
If a list of axes are given, it must match the number of partial dependence plots. If it does, the partial dependence plots will be place into those axes.
The example, plot_partial_dependence_visualization_api.py, is used to demonstrate how plot_partial_dependence how to use this axes based API.

Any other comments?

This PR removes the custom padding between subplots because the padding depends on the size of the figure. When using matplotlib's constrained_layout:

fig, ax = plt.subplots(constrained_layout=True)
disp = plot_partial_dependence(..., ax=ax)

The labels of the plots do not overlap. This can not be used in our examples because we are support matplotlib 1.5.1 (which does not have constrained_layout).

…ial_dep

amueller · 2019-08-13T21:54:10Z

could we use tight layout in the examples?

amueller · 2019-08-13T21:56:35Z

I think the possibility of providing a single axes and stealing or providing the right number of axes and reusing should be explained in the dev docs where you explained the display API

glemaitre

It generally looks good.

examples/inspection/plot_partial_dependence.py

glemaitre · 2019-08-14T11:29:04Z

examples/plot_partial_dependence_visualization_api.py

+The :func:`~sklearn.inspection.plot_partial_dependence` function returns a
+:class:`~sklearn.inspection.PartialDependenceDisplay` object that can be used
+for plotting without needing to recalculate the partial dependence. In this
+example we should how to plot partial dependence plots and quickly customize


Suggested change

example we should how to plot partial dependence plots and quickly customize

example, we show how to plot partial dependence plots and how to quickly customize

glemaitre · 2019-08-14T11:29:33Z

examples/plot_partial_dependence_visualization_api.py

+:class:`~sklearn.inspection.PartialDependenceDisplay` object that can be used
+for plotting without needing to recalculate the partial dependence. In this
+example we should how to plot partial dependence plots and quickly customize
+tht plot with the Visualization API.


Suggested change

tht plot with the Visualization API.

the plot with the Visualization API.

glemaitre · 2019-08-14T11:31:58Z

examples/plot_partial_dependence_visualization_api.py

+X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.1,
+                                                    random_state=0)
+
+hgbr = make_pipeline(StandardScaler(), HistGradientBoostingRegressor())


There is no need to use the StandardScaler. It will also speed-up the partial-dependence computation since we will use the recursive method included in the HistGradientBoostingRegressor which is not available when using a Pipeline.

glemaitre · 2019-08-14T11:34:09Z

examples/plot_partial_dependence_visualization_api.py

+# curves.
+#
+# One way to plot the curves is to place them in the same figure, with the
+# curves of each model on each row. First, we create a figure two axes with two


"a figure two axes" -> is it missing a "with"?

glemaitre · 2019-08-14T12:05:48Z

sklearn/inspection/partial_dependence.py

        For two-way partial dependence plots.

+    ax : Matplotlib axes, list of Matplotlib axes or None, (default=None)
+        By default,


Missing the rest of the sentence ;)

sklearn/inspection/partial_dependence.py

glemaitre · 2019-08-14T12:11:36Z

sklearn/inspection/tests/test_plot_partial_dependence.py

+from sklearn.inspection import plot_partial_dependence
+
+
+boston = load_boston()


it would be better to name it BOSTON.

Maybe this is time to use a fixture for the data here instead?

glemaitre · 2019-08-14T12:13:19Z

sklearn/inspection/tests/test_plot_partial_dependence.py

+    # Test partial dependence plot function on multi-output input.
+    X, y = multioutput_regression_data
+    clf = LinearRegression()
+    clf.fit(X, y)


are you using clf?

glemaitre

Ups but not to be approved yet :) wrong button

amueller · 2019-08-14T16:37:00Z

doc/developers/contributing.rst

@@ -1724,4 +1724,39 @@ attributes:
       return viz.plot(ax=ax, name=name, **kwargs)
 ```


Suggested change

```

amueller · 2019-08-14T16:40:39Z

doc/developers/contributing.rst

+the axes on the grid. Furthermore, the matplotlib Artists are stored in
+`lines_` and `contours_` where the key is the position on the grid. When a list
+of axes is passsed in, the `axes_`, `lines_`, and `contours_` keys is single
+int corresponding to the position on the passed in list of axes.


Can you link to the advanced partial dependence example here?

amueller · 2019-08-14T16:41:31Z

examples/plot_partial_dependence_visualization_api.py

@@ -0,0 +1,110 @@
+"""
+=========================================
+Partial Dependence with Visualization API


Advanced plotting with partial dependence?

Any partial dependence plot uses the visualization API by definition, right?

Love the example btw!

amueller · 2019-08-14T17:53:45Z

examples/plot_partial_dependence_visualization_api.py

+# curves.
+#
+# One way to plot the curves is to place them in the same figure, with the
+# curves of each model on each row. First, we create a figure two axes with two


Suggested change

# curves of each model on each row. First, we create a figure two axes with two

# curves of each model on each row. First, we create a figure with two axes within two

amueller

We might need to think about the data structures for self.axes_ a bit more. Or at least you should clarify.

amueller · 2019-08-14T17:54:57Z

examples/plot_partial_dependence_visualization_api.py

+# :func:`~sklearn.inspection.PartialDependenceDisplay.plot` functions of
+# `hgbr_disp` and `mlp_disp`. The plot funciton will plot the two curves in the
+# space by the passed in axes. The resulting plot places the histogram gradient
+# boosting partial dependence curves on top of the multi-layer perceptron


"places the gbrt plots in the first row and the mlp plots in the second row"?

amueller · 2019-08-14T17:56:07Z

examples/plot_partial_dependence_visualization_api.py

+# Another way to compare the curves is to plot them on top of each other. Here
+# we create a figure with one row and two columns. The axes are passed into the
+# :func:`~sklearn.inspection.PartialDependenceDisplay.plot` function as list,
+# which will plot the partial dependence curves of each model on the same axes.


Maybe add a sentence saying that the length of axes must be equal to the number of features we plot?

amueller · 2019-08-14T17:57:56Z

sklearn/inspection/_plot/partial_dependence.py

+    ----------
+    bounding_ax_ : matplotlib Axes or None
+        If `ax` is an axes or None, the `bounding_ax_` is the axes where the
+        grid of partial dependence plots are drawn. If `ax` is  list of axes,


Suggested change

grid of partial dependence plots are drawn. If `ax` is list of axes,

grid of partial dependence plots are drawn. If `ax` is a list of axes,

amueller · 2019-08-14T17:58:53Z

sklearn/inspection/_plot/partial_dependence.py

+        and jth column. If `ax` is a list of axes, `axes_[i]` is the ith item
+        in `ax`.
+
+    lines_ : matplotlib Artists


is this a dict? a list? a numpy array?

amueller · 2019-08-14T17:59:15Z

sklearn/inspection/_plot/partial_dependence.py

+        grid of partial dependence plots are drawn. If `ax` is  list of axes,
+        `bounding_ax_` is None.
+
+    axes_ : dict of matplotlib Axes


how can a dict have [i, j]?

It's a little weird that the shape of the attributes depends on whether we passed a list of axes or not, but I don't see a nice way around that right now.

amueller · 2019-08-14T18:01:45Z

sklearn/inspection/_plot/partial_dependence.py

+            given, it is treated as a bounding axes and a grid of partial
+            depdendence plots will be drawn on that top of it. If a list of
+            axes are passed, the partial dependence plots will be drawn on
+            those axes. By default, a single bounding axes is created and


remove single here?

amueller · 2019-08-14T18:03:30Z

sklearn/inspection/_plot/partial_dependence.py

+            those axes. By default, a single bounding axes is created and
+            treated as the single axes case.
+
+        n_cols : int, default=3


Honestly, I hate the way this is currently done and we should be doing what dabl is doing or https://github.com/matplotlib/grid-strategy is doing but that's for another day lol.

amueller · 2019-08-14T18:08:21Z

sklearn/inspection/_plot/partial_dependence.py

+            A figure object onto which the plots will be drawn, after the
+            figure has been cleared. By default, a new one is created.
+
+            .. deprecated:: 0.22


I agree. We need it for backward-compatibility though. We could make it private. I'm not sure if that's less weird?

sklearn/inspection/_plot/partial_dependence.py

+            if fig is None:
+                _, ax = plt.subplots()
+            else:
+                ax = fig.add_subplot(111)


I hate the matlab syntax. I would use either add_subplot(1, 1, 1) or gca?

amueller · 2019-08-14T18:14:31Z

examples/plot_partial_dependence_visualization_api.py

+# which will plot the partial dependence curves of each model on the same axes.
+
+# sphinx_gallery_thumbnail_number = 4
+fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(12, 6))


I feel like it would be more natural to just call hgbr_disp with the default arguments and use the generated axes to pass to mlp_disp.
Right now the API might require us to do a ravel or something? But that's something that we should support. Maybe we should be using OrderedDicts?

…ial_dep

thomasjpfan · 2019-08-16T15:31:18Z

API issues maybe?

https://nbviewer.jupyter.org/gist/thomasjpfan/54b17a41e81c1303f4dd0ce4cc539a2a

amueller · 2019-08-16T19:34:19Z

sklearn/inspection/_plot/partial_dependence.py

@@ -23,16 +23,16 @@ class PartialDependenceDisplay:

    Parameters
    ----------
-    pd_results : list of (ndarray, ndarray)
+    pd_results : (ndarray, ndarray) list


list of tuples?

thomasjpfan · 2019-08-16T19:48:10Z

If a user passes in a list or an ndarray of one dimension, axes_ will be one dimension and the len(features)==ax.shape[0] can be checked.

glemaitre · 2019-08-19T14:16:36Z

sklearn/inspection/_plot/partial_dependence.py

        Results of `sklearn.inspection.partial_dependence` for ``features``.
        Each tuple corresponds to a (averaged_predictions, grid).

-    features : list of {(int, ), (int, int)}
+    features {(int, ), (int, int)} list


Did I miss a discussion about this style?

…ial_dep

NicolasHug

Great work!

Made a bunch of comments. I only looked at the docstrings and the examples so far.

I don't think we really need the _plot folder, why not just have partial_dependence_display.py? If we must have _plot then it would make sense to me that plot_partial_dependence() is there. On top of that we now have 3 files named partial_dependence.py :/
About the deciles: (I'm literally discovering that now, shame on me), would you mind adding a small comment in the docstring of plot_partial_depence, e.g.
"The deciles of the feature values will be shown with tick marks on the x-axis "
With the current API, is it easy to extend it in the future by allowing e.g. 3d plots instead of contours for 2way pdps?

NicolasHug · 2019-09-03T17:45:50Z

doc/developers/plotting.rst

+
+When a single axes is passed in, that axes defines a space for the multiple
+axes to be placed. In this case, matplotlib's
+`gridspec.GridSpecFromSubplotSpec` can be used to split up the space::


I'm a bit confused. Is this a suggestion of what the internals of plot_* could/should be? If so, let's say it ;)

NicolasHug · 2019-09-03T17:46:37Z

doc/developers/plotting.rst

+By default, the `ax` keyworld in `plot` is `None`. In this case, the single
+axes is created and the gridspec api is used to create the regions to plot in.
+
+For example, :func:`~sklearn.inspection.plot_partial_dependence` plots multiple


I'd suggest "See for example [...] which plots multiple..."

NicolasHug · 2019-09-03T17:47:33Z

doc/developers/plotting.rst

+axes is created and the gridspec api is used to create the regions to plot in.
+
+For example, :func:`~sklearn.inspection.plot_partial_dependence` plots multiple
+lines and contours using this API. The axes that is passed in or created that


The axes that is passed in or created that defines the space

That sounds a bit odd to me

NicolasHug · 2019-09-03T17:49:26Z

doc/developers/plotting.rst

+the grid. Positions that are not used are set to `None`. Furthermore, the
+matplotlib Artists are stored in `lines_` and `contours_` where the key is the
+position on the grid. When a list of axes is passsed in, the `axes_`, `lines_`,
+and `contours_` keys is single int corresponding to the position on the passed


is single int -> "are integers" ?

on -> of ??

I find that whole sentence a bit hard to read.

NicolasHug · 2019-09-03T17:49:48Z

doc/developers/plotting.rst

+and `contours_` keys is single int corresponding to the position on the passed
+in list of axes. 
+
+Read more in :ref:`sphx_glr_auto_examples_plot_roc_curve_visualization_api.py`


should this be in the above section?

NicolasHug · 2019-09-03T18:34:02Z

examples/plot_partial_dependence_visualization_api.py

+# Plotting partial dependence of the two models independently
+# ===========================================================
+#
+# Next, we plot a partial dependence curves for features "LSTAT" and "RM" for


Suggested change

# Next, we plot a partial dependence curves for features "LSTAT" and "RM" for

# Next, we plot a partial dependence curves for features "LSTAT" and "RM"

NicolasHug · 2019-09-03T18:36:07Z

examples/plot_partial_dependence_visualization_api.py

+
+##############################################################################
+# The partial depdendence curves can be plotted for the multi-layer perceptron.
+# In this case `line_kw` was passed to


Suggested change

# In this case `line_kw` was passed to

# In this case `line_kw` is passed to

NicolasHug · 2019-09-03T18:36:13Z

examples/plot_partial_dependence_visualization_api.py

+##############################################################################
+# The partial depdendence curves can be plotted for the multi-layer perceptron.
+# In this case `line_kw` was passed to
+# `~sklearn.inspection.plot_partial_dependence` to change the color of the


does not link

NicolasHug · 2019-09-03T18:36:26Z

examples/plot_partial_dependence_visualization_api.py

+# The partial depdendence curves can be plotted for the multi-layer perceptron.
+# In this case `line_kw` was passed to
+# `~sklearn.inspection.plot_partial_dependence` to change the color of the
+# curve and `n_cols` was set to 1 to set the number of columns to 1.


Suggested change

# curve and `n_cols` was set to 1 to set the number of columns to 1.

# curve and `n_cols` is set to 1.

NicolasHug · 2019-09-03T18:44:37Z

examples/plot_partial_dependence_visualization_api.py

+#
+# One way to plot the curves is to place them in the same figure, with the
+# curves of each model on each row. First, we create a figure with two axes
+# within two rows and one column. The two axes are passed to the


I like the example but it's pretty condensed and it combines multiple functionalities at once so it's hard to understand how things interact independently.

For example it's a bit confusing at first how you can end up with a 2 by 2 grid while you only asked for 1 column.

I think it's missing a very simple plot with just 1 feature where you combine the curves of the 2 models. Then

I added a "Plotting partial dependence for one feature" section at the end. With the current API, there are two ways to plot a single feature:

Option 1

tree_disp = plot_partial_dependence(tree, X, ["LSTAT"], feature_names=feature_names) mlp_disp = plot_partial_dependence(mlp, X, ["LSTAT"], feature_names=feature_names, ax=tree_disp.axes_, line_kw={"c": "red"})

Option 2

_, ax = plt.subplots() tree_disp = plot_partial_dependence(tree, X, ["LSTAT"], feature_names=feature_names, ax=[ax]) mlp_disp = plot_partial_dependence(mlp, X, ["LSTAT"], feature_names=feature_names, ax=[ax], line_kw={"c": "red"})

For the example, I went with option 1. The "nicer" way to do this is:

Possible option that is being disallowed.

_, ax = plt.subplots() tree_disp = plot_partial_dependence(tree, X, ["LSTAT"], feature_names=feature_names, ax=ax) mlp_disp = plot_partial_dependence(mlp, X, ["LSTAT"], feature_names=feature_names, ax=ax, line_kw={"c": "red"})

The first call will call ax.set_visible(False) to denote that the space has been used. The second call will see that the axes is not visible and raise an error. We can technically support this "single feature" and "single axes" case, which I think will add another layer of complexity to the API i.e. "If ax is a single axes and len(features) == 1, then we behave differently"

NicolasHug · 2019-09-04T15:10:01Z

sklearn/inspection/_plot/partial_dependence.py

+            A figure object onto which the plots will be drawn, after the
+            figure has been cleared. By default, a new one is created.
+
+            .. deprecated:: 0.22


Should we just not publicly document it then?

NicolasHug · 2019-09-04T15:11:48Z

doc/developers/plotting.rst

+
+Scikit-learn defines a simple API for creating visualizations for machine
+learning. The key features of this API is to run calculations once and to have
+the flexibility to adjust the visualizations after the fact.


To explain the purpose of this doc we can add something like

This section is intended for developers who whish to develop or maintain plotting tools. For usage, users should refer to the user guide <link>

thomasjpfan · 2019-09-04T16:24:41Z

I don't think we really need the _plot folder,

I have trying to keep using this _plot folder to place the display and plot function into its namespace and only import it in sklearn.inspection. This can not be done here because plot_partial_dependence is already defined publicly in sklearn.inspeciton.partial_dependence. So for only partial_dependence it can be justified to create a _partial_dependence_display.py file.

NicolasHug · 2019-09-04T16:36:29Z

I won't argue more but I find that having a module with a single file isn't necessary (esp. since it creates other kinds of inconsistencies), and having 3 different files named partial_dependence.py is confusing.

thomasjpfan · 2019-09-05T15:53:09Z

With the current API, is it easy to extend it in the future by allowing e.g. 3d plots instead of contours for 2way pdps?

I have been avoiding 3-d plots. They... add another dimension to plotting. I think current plot_partial_dependence API is kind of bloated with its 2-way partial dependence, I would not want to add more to it.

…ial_dep

doc/modules/classes.rst

…ial_dep

sklearn/inspection/partial_dependence.py

glemaitre · 2019-09-16T12:42:01Z

sklearn/inspection/partial_dependence.py

+            A figure object onto which the plots will be drawn, after the
+            figure has been cleared. By default, a new one is created.
+
+            .. deprecated:: 0.22


Suggested change

.. deprecated:: 0.22

.. deprecated:: 0.22

``fig`` will be removed in 0.24.

glemaitre · 2019-09-16T12:42:24Z

sklearn/inspection/partial_dependence.py

+            Dict with keywords passed to the `matplotlib.pyplot.contourf`
+            call for two-way partial dependence plots.
+
+        fig : Matplotlib figure object, optional (default=None)


Suggested change

fig : Matplotlib figure object, optional (default=None)

fig : Matplotlib figure object, default=None

Is there a way to not introduce a parameter which is directly deprecated.

Updated PR to not need to add a fig just to deprecate it.

glemaitre · 2019-09-16T12:45:35Z

examples/plot_partial_dependence_visualization_api.py

+:class:`~sklearn.inspection.PartialDependenceDisplay` object that can be used
+for plotting without needing to recalculate the partial dependence. In this
+example, we show how to plot partial dependence plots and how to quickly
+customize the plot with the Visualization API.


Suggested change

customize the plot with the Visualization API.

customize the plot with the visualization API.

glemaitre · 2019-09-16T12:53:54Z

sklearn/inspection/partial_dependence.py

        For two-way partial dependence plots.

+    ax : Matplotlib axes or array-like of Matplotlib axes, default=None
+        - If a single axes is passed in, it is treated as a bounding axes


Suggested change

- If a single axes is passed in, it is treated as a bounding axes

- If a single axis is passed in, it is treated as a bounding axes

glemaitre · 2019-09-16T12:54:14Z

sklearn/inspection/partial_dependence.py

+            and a grid of partial depedendence plots will be drawn within
+            these bounds. The `n_cols` parameter controls the number of
+            columns in the grid.
+        - If a array-like of axes are passed in, the partial dependence


Suggested change

- If a array-like of axes are passed in, the partial dependence

- If an array-like of axes are passed in, the partial dependence

glemaitre · 2019-09-16T12:54:47Z

sklearn/inspection/partial_dependence.py

+            columns in the grid.
+        - If a array-like of axes are passed in, the partial dependence
+            plots will be drawn directly into these axes.
+        - If `None`, a figure and a bounding axes is created and treated


Suggested change

- If `None`, a figure and a bounding axes is created and treated

- If `None`, a figure and a bounding axis is created and treated

glemaitre · 2019-09-16T12:54:55Z

sklearn/inspection/partial_dependence.py

+        - If a array-like of axes are passed in, the partial dependence
+            plots will be drawn directly into these axes.
+        - If `None`, a figure and a bounding axes is created and treated
+            as the single axes case.


Suggested change

as the single axes case.

as the single axis case.

glemaitre · 2019-09-16T12:56:23Z

sklearn/inspection/partial_dependence.py

+    display = PartialDependenceDisplay(pd_results, features, feature_names,
+                                       target_idx, pdp_lim, deciles)
+    return display.plot(ax=ax, n_cols=n_cols, line_kw=line_kw,
+                        contour_kw=contour_kw, fig=fig)


Regarding the introduction of fig

…ial_dep

glemaitre · 2019-09-20T09:03:47Z

LGTM Thanks @thomasjpfan

thomasjpfan added 7 commits August 10, 2019 11:57

WIP

fa93505

Merge remote-tracking branch 'upstream/master' into plotting_api_part…

d1dc257

…ial_dep

WIP Adding tests

1dac62c

WIP Doc

bc45123

CLN Creates _plot folder

9d67f6a

CLN Adds _plot into setup

5129289

CLN Fixes bugs

6f5407e

glemaitre approved these changes Aug 14, 2019

View reviewed changes

glemaitre requested changes Aug 14, 2019

View reviewed changes

DOC Adds dev docs

8bc2433

amueller reviewed Aug 14, 2019

View reviewed changes

thomasjpfan added 5 commits August 14, 2019 16:00

ENH Updates API

4aa273c

Merge remote-tracking branch 'upstream/master' into plotting_api_part…

5263235

…ial_dep

CLN Address comments

045aa8f

TST Removes check

75f8ed8

DOC Address comments

5fddc5f

thomasjpfan added 4 commits August 16, 2019 14:29

DOC Uses normal decision tree in example

396f1cb

CLN Refactor array-like

bd121c3

STY Minor

422992e

DOC Update

ca532d7

amueller reviewed Aug 16, 2019

View reviewed changes

amueller approved these changes Aug 16, 2019

View reviewed changes

glemaitre reviewed Aug 19, 2019

View reviewed changes

thomasjpfan added 4 commits August 26, 2019 08:00

DOC Update docstring

9b916cf

Merge remote-tracking branch 'upstream/master' into plotting_api_part…

ec47bd8

…ial_dep

DOC Adds new plotting api section

170a1b5

DOC Uses :: syntax

9ad9477

NicolasHug reviewed Sep 3, 2019

View reviewed changes

NicolasHug reviewed Sep 4, 2019

View reviewed changes

thomasjpfan added 3 commits September 4, 2019 15:06

CLN Moves files around

cab552a

DOC Updates plotting.rst

0f5dee1

DOC Address comments

e7965d9

thomasjpfan added 6 commits September 5, 2019 11:53

CLN Address comments

24ee8b0

CLN Address comments

d339328

Merge remote-tracking branch 'upstream/master' into plotting_api_part…

f7e946d

…ial_dep

DOC More links

480bcef

DOC More words

e9e5d3a

BUG Fix setup file

4b50a8b

glemaitre reviewed Sep 9, 2019

View reviewed changes

doc/modules/classes.rst Show resolved Hide resolved

thomasjpfan added 3 commits September 9, 2019 13:19

DOC Fix

f77f5d1

REV Remove unrelated commit

612a80e

Merge remote-tracking branch 'upstream/master' into plotting_api_part…

a8182d8

…ial_dep

glemaitre reviewed Sep 16, 2019

View reviewed changes

thomasjpfan added 2 commits September 19, 2019 12:46

Merge remote-tracking branch 'upstream/master' into plotting_api_part…

6c3a6c9

…ial_dep

CLN Address @glemaitre comments

42e0c9b

glemaitre merged commit 97185ec into scikit-learn:master Sep 20, 2019

glemaitre approved these changes Sep 20, 2019

View reviewed changes

jnothman mentioned this pull request Oct 28, 2019

Missing from what's new v0.22 #15384

Closed

8 tasks

	example we should how to plot partial dependence plots and quickly customize
	example, we show how to plot partial dependence plots and how to quickly customize

	tht plot with the Visualization API.
	the plot with the Visualization API.

		from sklearn.inspection import plot_partial_dependence


		boston = load_boston()

		@@ -1724,4 +1724,39 @@ attributes:
		return viz.plot(ax=ax, name=name, **kwargs)
		```

	# curves of each model on each row. First, we create a figure two axes with two
	# curves of each model on each row. First, we create a figure with two axes within two

	grid of partial dependence plots are drawn. If `ax` is list of axes,
	grid of partial dependence plots are drawn. If `ax` is a list of axes,

	# Next, we plot a partial dependence curves for features "LSTAT" and "RM" for
	# Next, we plot a partial dependence curves for features "LSTAT" and "RM"

	# In this case `line_kw` was passed to
	# In this case `line_kw` is passed to

	# curve and `n_cols` was set to 1 to set the number of columns to 1.
	# curve and `n_cols` is set to 1.

	.. deprecated:: 0.22
	.. deprecated:: 0.22
	``fig`` will be removed in 0.24.

	fig : Matplotlib figure object, optional (default=None)
	fig : Matplotlib figure object, default=None

	customize the plot with the Visualization API.
	customize the plot with the visualization API.

	- If a single axes is passed in, it is treated as a bounding axes
	- If a single axis is passed in, it is treated as a bounding axes

	- If a array-like of axes are passed in, the partial dependence
	- If an array-like of axes are passed in, the partial dependence

	- If `None`, a figure and a bounding axes is created and treated
	- If `None`, a figure and a bounding axis is created and treated

Uh oh!

[MRG] Adds Plotting API to Partial Dependence #14646

[MRG] Adds Plotting API to Partial Dependence #14646

Uh oh!

Conversation

Uh oh!

Reference Issues/PRs

What does this implement/fix? Explain your changes.

Any other comments?

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

B41A Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!