UnrealNAS: Can We Search Neural Architectures with Unreal Data?
This work addresses the problem of data scarcity in NAS for applications with limited accessible data, offering a potential solution by leveraging unreal data, though it appears incremental as it builds on prior analysis of label necessity.
The paper investigates whether neural architecture search (NAS) can be effective using unreal data instead of real labeled data, and finds that architectures searched on constructed unreal datasets achieve promising results comparable to those from conventional NAS with real data, as shown in experiments on CIFAR, ImageNet, and CheXpert.
Neural architecture search (NAS) has shown great success in the automatic design of deep neural networks (DNNs). However, the best way to use data to search network architectures is still unclear and under exploration. Previous work has analyzed the necessity of having ground-truth labels in NAS and inspired broad interest. In this work, we take a further step to question whether real data is necessary for NAS to be effective. The answer to this question is important for applications with limited amount of accessible data, and can help people improve NAS by leveraging the extra flexibility of data generation. To explore if NAS needs real data, we construct three types of unreal datasets using: 1) randomly labeled real images; 2) generated images and labels; and 3) generated Gaussian noise with random labels. These datasets facilitate to analyze the generalization and expressivity of the searched architectures. We study the performance of architectures searched on these constructed datasets using popular differentiable NAS methods. Extensive experiments on CIFAR, ImageNet and CheXpert show that the searched architectures can achieve promising results compared with those derived from the conventional NAS pipeline with real labeled data, suggesting the feasibility of performing NAS with unreal data.