# On Network Design Spaces for Visual Recognition

## TL;DR

This paper introduces a new statistical methodology for comparing neural network design spaces in visual recognition, providing more comprehensive insights into their differences and similarities, especially in neural architecture search.

## Contribution

It proposes a distribution-based comparison paradigm for network design spaces, enhancing the robustness of architecture evaluation beyond traditional point estimates.

## Key findings

- Significant statistical differences exist between recent NAS design spaces.
- Standard models like ResNeXt can be comparable to NAS design spaces.
- Distribution estimates offer a more complete view of the design landscape.

## Abstract

Over the past several years progress in designing better neural network architectures for visual recognition has been substantial. To help sustain this rate of progress, in this work we propose to reexamine the methodology for comparing network architectures. In particular, we introduce a new comparison paradigm of distribution estimates, in which network design spaces are compared by applying statistical techniques to populations of sampled models, while controlling for confounding factors like network complexity. Compared to current methodologies of comparing point and curve estimates of model families, distribution estimates paint a more complete picture of the entire design landscape. As a case study, we examine design spaces used in neural architecture search (NAS). We find significant statistical differences between recent NAS design space variants that have been largely overlooked. Furthermore, our analysis reveals that the design spaces for standard model families like ResNeXt can be comparable to the more complex ones used in recent NAS work. We hope these insights into distribution analysis will enable more robust progress toward discovering better networks for visual recognition.