Semi-Supervised Neural Architecture Search

TL;DR

SemiNAS introduces a semi-supervised approach to neural architecture search that leverages unlabeled architectures to reduce computational costs and improve accuracy, demonstrating effectiveness on multiple benchmarks and tasks.

Contribution

It proposes SemiNAS, a semi-supervised NAS method that uses unlabeled architectures to enhance predictor training and reduce search costs.

Findings

Achieves comparable accuracy with only 1/7 labeled data on NASBench-101.

Outperforms baselines with higher accuracy at the same computational cost.

Attains 94.02% test accuracy on NASBench-101 and 23.5% top-1 error on ImageNet.

Abstract

Neural architecture search (NAS) relies on a good controller to generate better architectures or predict the accuracy of given architectures. However, training the controller requires both abundant and high-quality pairs of architectures and their accuracy, while it is costly to evaluate an architecture and obtain its accuracy. In this paper, we propose SemiNAS, a semi-supervised NAS approach that leverages numerous unlabeled architectures (without evaluation and thus nearly no cost). Specifically, SemiNAS 1) trains an initial accuracy predictor with a small set of architecture-accuracy data pairs; 2) uses the trained accuracy predictor to predict the accuracy of large amount of architectures (without evaluation); and 3) adds the generated data pairs to the original data to further improve the predictor. The trained accuracy predictor can be applied to various NAS algorithms by predicting the accuracy of candidate architectures for them. SemiNAS has two advantages: 1) It reduces the computational cost under the same accuracy guarantee. On NASBench-101 benchmark dataset, it achieves comparable accuracy with gradient-based method while using only 1/7 architecture-accuracy pairs. 2) It achieves higher accuracy under the same computational cost. It achieves 94.02% test accuracy on NASBench-101, outperforming all the baselines when using the same number of architectures. On ImageNet, it achieves 23.5% top-1 error rate (under 600M FLOPS constraint) using 4 GPU-days for search. We further apply it to LJSpeech text to speech task and it achieves 97% intelligibility rate in the low-resource setting and 15% test error rate in the robustness setting, with 9%, 7% improvements over the baseline respectively.