Scalable Interpretability via PolynomialsDownload PDF

Published: 31 Oct 2022, Last Modified: 12 Mar 2024NeurIPS 2022 AcceptReaders: Everyone
Keywords: interpretability, explainability, trustworthy AI, polynomials, generalized additive models, interpretable machine learning
TL;DR: Second degree polynomials can be used as drop-in replacements for DNNs on most tabular and processed image datasets for interpretability with no loss in performance.
Abstract: Generalized Additive Models (GAMs) have quickly become the leading choice for interpretable machine learning. However, unlike uninterpretable methods such as DNNs, they lack expressive power and easy scalability, and are hence not a feasible alternative for real-world tasks. We present a new class of GAMs that use tensor rank decompositions of polynomials to learn powerful, {\em inherently-interpretable} models. Our approach, titled Scalable Polynomial Additive Models (SPAM) is effortlessly scalable and models {\em all} higher-order feature interactions without a combinatorial parameter explosion. SPAM outperforms all current interpretable approaches, and matches DNN/XGBoost performance on a series of real-world benchmarks with up to hundreds of thousands of features. We demonstrate by human subject evaluations that SPAMs are demonstrably more interpretable in practice, and are hence an effortless replacement for DNNs for creating interpretable and high-performance systems suitable for large-scale machine learning. Source code is available at \href{}{\ttfamily}.
Supplementary Material: pdf
Community Implementations: [![CatalyzeX](/images/catalyzex_icon.svg) 1 code implementation](
12 Replies