Accuracy, Interpretability, and Differential Privacy via Explainable Boosting

TL;DR

This paper demonstrates that integrating differential privacy with Explainable Boosting Machines achieves high accuracy and interpretability, enabling privacy-preserving, understandable, and editable machine learning models.

Contribution

The paper introduces DP-EBM, a method combining differential privacy with EBMs, maintaining accuracy and interpretability while allowing post-training model editing.

Findings

DP-EBM achieves state-of-the-art accuracy with strong privacy guarantees.

Models provide exact global and local interpretability.

Models can be edited after training without privacy loss.

Abstract

We show that adding differential privacy to Explainable Boosting Machines (EBMs), a recent method for training interpretable ML models, yields state-of-the-art accuracy while protecting privacy. Our experiments on multiple classification and regression datasets show that DP-EBM models suffer surprisingly little accuracy loss even with strong differential privacy guarantees. In addition to high accuracy, two other benefits of applying DP to EBMs are: a) trained models provide exact global and local interpretability, which is often important in settings where differential privacy is needed; and b) the models can be edited after training without loss of privacy to correct errors which DP noise may have introduced.