HyperSHAP: Shapley Values and Interactions for Explaining Hyperparameter Optimization
Marcel Wever, Maximilian Muschalik, Fabian Fumagalli, Marius Lindauer
TL;DR
HyperSHAP presents a game-theoretic, post-hoc explainability framework for hyperparameter optimization that uses Shapley values and interactions to dissect performance across hyperparameters. It defines five explanation games (Ablation, Sensitivity, Tunability, Optimizer Bias, and Multi-Dataset extensions) and derives both local and global explanations via $\phi^{\text{SV}}$ and $\Phi_k$, enabling principled analysis of parameter contributions and their interactions. Compared to prior methods like fANOVA, HyperSHAP emphasizes actionable tunability insights and detector bias in optimizers, supported by experiments on multiple HPO benchmarks and surrogate models. The findings indicate higher-order interactions exist but are mostly captured by up to third-order terms, and HyperSHAP provides scalable, interpretable guidance for selecting hyperparameters to tune and diagnosing optimizer behavior.
Abstract
Hyperparameter optimization (HPO) is a crucial step in achieving strong predictive performance. Yet, the impact of individual hyperparameters on model generalization is highly context-dependent, prohibiting a one-size-fits-all solution and requiring opaque HPO methods to find optimal configurations. However, the black-box nature of most HPO methods undermines user trust and discourages adoption. To address this, we propose a game-theoretic explainability framework for HPO based on Shapley values and interactions. Our approach provides an additive decomposition of a performance measure across hyperparameters, enabling local and global explanations of hyperparameters' contributions and their interactions. The framework, named HyperSHAP, offers insights into ablation studies, the tunability of learning algorithms, and optimizer behavior across different hyperparameter spaces. We demonstrate HyperSHAP's capabilities on various HPO benchmarks to analyze the interaction structure of the corresponding HPO problems, demonstrating its broad applicability and actionable insights for improving HPO.