Beyond Sparsity: Tree Regularization of Deep Models for Interpretability

TL;DR

This paper introduces a tree regularization technique for deep models that enhances interpretability by enabling models to be approximated by small decision trees, maintaining accuracy while improving human understanding.

Contribution

It proposes a novel tree regularization method for deep models, improving interpretability without sacrificing predictive performance.

Findings

Models are more interpretable and easier for humans to simulate.

Tree regularization maintains high accuracy on medical and toy datasets.

Outperforms L1 and L2 penalties in interpretability without accuracy loss.

Abstract

The lack of interpretability remains a key barrier to the adoption of deep models in many applications. In this work, we explicitly regularize deep models so human users might step through the process behind their predictions in little time. Specifically, we train deep time-series models so their class-probability predictions have high accuracy while being closely modeled by decision trees with few nodes. Using intuitive toy examples as well as medical tasks for treating sepsis and HIV, we demonstrate that this new tree regularization yields models that are easier for humans to simulate than simpler L1 or L2 penalties without sacrificing predictive power.