Neural-ANOVA: Analytical Model Decomposition using Automatic Integration
This work addresses the need for interpretability in neural networks by providing a systematic decomposition method, though it appears incremental as it builds on existing ANOVA techniques.
The paper tackled the problem of decomposing neural networks into lower-order models to understand interaction effects, and introduced Neural-ANOVA, which enables fast analytical evaluation of integrals for this decomposition, with numerical experiments showing insights into approximation properties compared to other regression methods.
The analysis of variance (ANOVA) decomposition offers a systematic method to understand the interaction effects that contribute to a specific decision output. In this paper we introduce Neural-ANOVA, an approach to decompose neural networks into the sum of lower-order models using the functional ANOVA decomposition. Our approach formulates a learning problem, which enables fast analytical evaluation of integrals over subspaces that appear in the calculation of the ANOVA decomposition. Finally, we conduct numerical experiments to provide insights into the approximation properties compared to other regression approaches from the literature.