ReLU Code Space: A Basis for Rating Network Quality Besides Accuracy

TL;DR

This paper introduces a novel ReLU code space with a truncated Hamming distance that correlates network activation patterns with safety and robustness, enabling new insights beyond accuracy.

Contribution

It establishes an isometry between activation codes and input space regions, facilitating analysis of network robustness and confidence.

Findings

Code space relates to safety and robustness

Efficient adjacency computation between input regions

Information beyond accuracy is stored in codes

Abstract

We propose a new metric space of ReLU activation codes equipped with a truncated Hamming distance which establishes an isometry between its elements and polyhedral bodies in the input space which have recently been shown to be strongly related to safety, robustness, and confidence. This isometry allows the efficient computation of adjacency relations between the polyhedral bodies. Experiments on MNIST and CIFAR-10 indicate that information besides accuracy might be stored in the code space.