Intriguing properties of neural networks

TLDR

Deep neural networks are highly expressive models that achieve state‑of‑the‑art performance on speech and visual recognition tasks, but this expressiveness also leads to uninterpretable solutions with counter‑intuitive properties. The study reports two counter‑intuitive properties of deep neural networks. The authors generate imperceptible perturbations that maximize prediction error to induce misclassification. They find that high‑level units and their random linear combinations are indistinguishable, that semantic information resides in the space rather than individual units, that networks learn discontinuous mappings, and that the same imperceptible perturbation can misclassify inputs across independently trained models.

Abstract

Deep neural networks are highly expressive models that have recently achieved state of the art performance on speech and visual recognition tasks. While their expressiveness is the reason they succeed, it also causes them to learn uninterpretable solutions that could have counter-intuitive properties. In this paper we report two such properties. First, we find that there is no distinction between individual high level units and random linear combinations of high level units, according to various methods of unit analysis. It suggests that it is the space, rather than the individual units, that contains of the semantic information in the high layers of neural networks. Second, we find that deep neural networks learn input-output mappings that are fairly discontinuous to a significant extend. We can cause the network to misclassify an image by applying a certain imperceptible perturbation, which is found by maximizing the network's prediction error. In addition, the specific nature of these perturbations is not a random artifact of learning: the same perturbation can cause a different network, that was trained on a different subset of the dataset, to misclassify the same input.