ReLU Code Space: A Basis for Rating Network Quality Besides Accuracy
Machine Learning
2020-05-21 v1 Machine Learning
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.
Cite
@article{arxiv.2005.09903,
title = {ReLU Code Space: A Basis for Rating Network Quality Besides Accuracy},
author = {Natalia Shepeleva and Werner Zellinger and Michal Lewandowski and Bernhard Moser},
journal= {arXiv preprint arXiv:2005.09903},
year = {2020}
}
Comments
in ICLR 2020 Workshop on Neural Architecture Search (NAS 2020)