SmoothHess: ReLU Network Feature Interactions via Stein's Lemma
This provides a practical tool for interpretability in ReLU networks, addressing a specific bottleneck in feature interaction analysis, but is incremental as it builds on existing Hessian-based methods.
The paper tackled the challenge of estimating feature interactions in ReLU networks, which have zero Hessians almost everywhere, by proposing SmoothHess, a method using Stein's Lemma to estimate second-order interactions via Gaussian convolution, and demonstrated its superior performance on benchmark and medical datasets.
Several recent methods for interpretability model feature interactions by looking at the Hessian of a neural network. This poses a challenge for ReLU networks, which are piecewise-linear and thus have a zero Hessian almost everywhere. We propose SmoothHess, a method of estimating second-order interactions through Stein's Lemma. In particular, we estimate the Hessian of the network convolved with a Gaussian through an efficient sampling algorithm, requiring only network gradient calls. SmoothHess is applied post-hoc, requires no modifications to the ReLU network architecture, and the extent of smoothing can be controlled explicitly. We provide a non-asymptotic bound on the sample complexity of our estimation procedure. We validate the superior ability of SmoothHess to capture interactions on benchmark datasets and a real-world medical spirometry dataset.