Similarity of Neural Architectures using Adversarial Attack Transferability
This work addresses the need for scalable similarity measures in deep learning to guide architecture design, though it is incremental as it builds on existing adversarial attack concepts.
The paper tackled the problem of quantifying similarity between neural architectures by proposing Similarity by Attack Transferability (SAT), based on adversarial attack transferability, and applied it to 69 ImageNet classifiers, finding that model diversity improves ensemble performance and knowledge distillation under specific conditions.
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.