This page contains the demo code to plot the step functions shown in our AISTATS 2022 paper:
To plot sparse step functions, we need to do the following two steps:
Binarize all continuous features into {0, 1} binary features by creating thresholds. This is a preprocessing step. The binarization helper function is in binarization_utils.py
An example is given in the binarize_continuousData_demo notebook.
Apply fastSparse on the preprocessed binary features to produce sparse coefficients and apply the plot functions in plot_utils.py.
An example is given in the plot_stepFunction_demo notebook, which reproduces the FICO step functions shown in the paper. Or you can modify the notebook to tailor to plot step functions on your own data.