Scikit-learn

Topics: Python css html ML Go back

At the time of writing, I have a contract to contribute to scikit-learn. My work is part of a grant awarded to support the improvement of the evaluation and inspection of predictive models in scikit-learn. The grant was requested by Guillaume Lemaitre and is administered by NumFOCUS.

The initial phase of my contract was to improve the estimators’ HTML representation. I added a table to the representation that displays the estimators’ parameters and their values. Non-default parameters—those set by the user—are highlighted. A copy-paste button is available for each parameter name. The parameters table is hidden by default.

The benefit of this new feature, is that it is interactive and for the user it s much simpler to know what's in the estimator.

This contribution was featured in the 1.7 release highlights: scikit-learn release notes

The following partial image, taken from the release notes, shows blocks displaying the parameters tables. However, it is not interactive, as it is meant for illustration only.