PDD-SHAP: Fast Approximations for Shapley Values using Functional Decomposition

TL;DR

PDD-SHAP introduces a fast approximation method for Shapley values using ANOVA-based functional decomposition, enabling efficient explanations of black-box models especially on large datasets.

Contribution

The paper presents PDD-SHAP, a novel algorithm that significantly speeds up Shapley value computation through functional decomposition, reducing costs for large-scale explanations.

Findings

Achieves orders of magnitude faster Shapley value approximations

Reduces computational costs for large datasets

Maintains accuracy comparable to exact methods

Abstract

Because of their strong theoretical properties, Shapley values have become very popular as a way to explain predictions made by black box models. Unfortuately, most existing techniques to compute Shapley values are computationally very expensive. We propose PDD-SHAP, an algorithm that uses an ANOVA-based functional decomposition model to approximate the black-box model being explained. This allows us to calculate Shapley values orders of magnitude faster than existing methods for large datasets, significantly reducing the amortized cost of computing Shapley values when many predictions need to be explained.