Overlapping plots

[idomic]

When trying to print multiple plots via the same cell, the matrix and feature importance are overlapping.

i.e:

KNN = KNeighborsClassifier()
KNN.fit(X_train, y_train)
predictions= KNN.predict(X_test)
print(accuracy_score(y_test, predictions))
print(classification_report(y_test, predictions))
knn_cm = plot.confusion_matrix(y_test, predictions)
plot.feature_importances(rfc, top_n=5)

[edublancas]

Ah, this problem goes back a while due to an error I made when creating the first plots.

Our plotting methods have a ax argument that allows you to use a matplotlib.Axes object and add your plot there. The typical use case is someone plotting 4 plots in a 2 x 2 grid, and then passing the Axes objects to the plotting methods to fill the grid.

However, when a user does not pass the ax argument, we set a default, currently, we're setting it as:

ax = plt.gca()

This means "get current axes". The problem with this approach is that it'll re-use the same axes (that's why the plots overlap). I'm unsure what's matplotlib's rule for deciding when to create a new Axes object. Maybe if running it in a different cell will force creating a new axes?

I think what's best is to do:

_, ax = plt.subplots()

just like sklearn does

This will create a new axes object. I don't think this will have any impact for end-users, but once we make the change, we can check the rendered docs and examples to see if anything looks odd.

[edublancas]

found another instance of overlapping plots:

https://sklearn-evaluation.ploomber.io/en/latest/api/plot.html#confusionmatrix

[edublancas]

note: I fixed this for confusion matrix, this PR can be used as reference: #208

[WSShawn]

okay! I'll check it out!

[edublancas]

this has been fixed #211

[idomic]

Is there anything else or we cna close this?

[edublancas]

nothing else to do, we can close it