Similarity of Neural Architectures using Adversarial Attack Transferability

TL;DR

This paper introduces a new similarity measure for neural architectures based on adversarial attack transferability, providing insights into model diversity, ensemble performance, and knowledge distillation.

Contribution

We propose the SAT measure to quantify neural architecture similarity using adversarial transferability, enabling large-scale analysis of model relationships.

Findings

Model diversity improves ensemble performance.

Distinct architectures enhance knowledge distillation.

Neural architecture similarity correlates with transferability of adversarial attacks.

Abstract

In recent years, many deep neural architectures have been developed for image classification. Whether they are similar or dissimilar and what factors contribute to their (dis)similarities remains curious. To address this question, we aim to design a quantitative and scalable similarity measure between neural architectures. We propose Similarity by Attack Transferability (SAT) from the observation that adversarial attack transferability contains information related to input gradients and decision boundaries widely used to understand model behaviors. We conduct a large-scale analysis on 69 state-of-the-art ImageNet classifiers using our proposed similarity function to answer the question. Moreover, we observe neural architecture-related phenomena using model similarity that model diversity can lead to better performance on model ensembles and knowledge distillation under specific conditions. Our results provide insights into why developing diverse neural architectures with distinct components is necessary.