Searching the Search Space of Vision Transformer

TL;DR

This paper introduces a neural architecture search method that automates the design of vision transformers by exploring their search space, resulting in models that outperform existing architectures on multiple vision tasks.

Contribution

It proposes a novel search process for vision transformers that includes searching the architecture and its search space, guided by E-T Error, and provides design guidelines for effective models.

Findings

S3 models outperform Swin, DeiT, and ViT on ImageNet

S3 models show strong performance on object detection, segmentation, and VQA

The search process offers insights into effective vision transformer design

Abstract

Vision Transformer has shown great visual representation power in substantial vision tasks such as recognition and detection, and thus been attracting fast-growing efforts on manually designing more effective architectures. In this paper, we propose to use neural architecture search to automate this process, by searching not only the architecture but also the search space. The central idea is to gradually evolve different search dimensions guided by their E-T Error computed using a weight-sharing supernet. Moreover, we provide design guidelines of general vision transformers with extensive analysis according to the space searching process, which could promote the understanding of vision transformer. Remarkably, the searched models, named S3 (short for Searching the Search Space), from the searched space achieve superior performance to recently proposed models, such as Swin, DeiT and ViT, when evaluated on ImageNet. The effectiveness of S3 is also illustrated on object detection, semantic segmentation and visual question answering, demonstrating its generality to downstream vision and vision-language tasks. Code and models will be available at https://github.com/microsoft/Cream.