Interpretability

Monotonic Networks

A small package to make neural networks monotonic in any subset of their inputs (this works for individual neurons, too!).

MoDe: Controlling Classifier Bias

A regularization to make neural networks' output independent from certain features.