Scalable NAS with Factorizable Architectural Parameters

TL;DR

This paper introduces a scalable neural architecture search method that factorizes large operator sets into smaller subspaces, enabling efficient exploration of architectures with state-of-the-art results on CIFAR10 and ImageNet.

Contribution

It proposes a novel factorization approach to scale NAS search spaces efficiently, including searching for activation functions alongside traditional operators.

Findings

Achieved state-of-the-art performance on CIFAR10.

Achieved state-of-the-art performance on ImageNet.

Efficiently searched for effective activation functions.

Abstract

Neural Architecture Search (NAS) is an emerging topic in machine learning and computer vision. The fundamental ideology of NAS is using an automatic mechanism to replace manual designs for exploring powerful network architectures. One of the key factors of NAS is to scale-up the search space, e.g., increasing the number of operators, so that more possibilities are covered, but existing search algorithms often get lost in a large number of operators. For avoiding huge computing and competition among similar operators in the same pool, this paper presents a scalable algorithm by factorizing a large set of candidate operators into smaller subspaces. As a practical example, this allows us to search for effective activation functions along with the regular operators including convolution, pooling, skip-connect, etc. With a small increase in search costs and no extra costs in re-training, we find interesting architectures that were not explored before, and achieve state-of-the-art performance on CIFAR10 and ImageNet, two standard image classification benchmarks.