XNAS: Neural Architecture Search with Expert Advice

TL;DR

This paper presents XNAS, a neural architecture search method based on expert advice theory, which dynamically eliminates inferior architectures and achieves strong results on image classification tasks.

Contribution

It introduces a novel optimization approach for neural architecture search that minimizes regret and dynamically prunes architectures, outperforming previous methods.

Findings

Achieves 1.6% error on CIFAR-10

Attains 24% error on ImageNet with mobile settings

Sets new state-of-the-art results on three datasets

Abstract

This paper introduces a novel optimization method for differential neural architecture search, based on the theory of prediction with expert advice. Its optimization criterion is well fitted for an architecture-selection, i.e., it minimizes the regret incurred by a sub-optimal selection of operations. Unlike previous search relaxations, that require hard pruning of architectures, our method is designed to dynamically wipe out inferior architectures and enhance superior ones. It achieves an optimal worst-case regret bound and suggests the use of multiple learning-rates, based on the amount of information carried by the backward gradients. Experiments show that our algorithm achieves a strong performance over several image classification datasets. Specifically, it obtains an error rate of 1.6% for CIFAR-10, 24% for ImageNet under mobile settings, and achieves state-of-the-art results on three additional datasets.