Few-shot Neural Architecture Search

Efficient evaluation of a network architecture drawn from a large search space remains a key challenge in Neural Architecture Search (NAS). Vanilla NAS evaluates each architecture by training from scratch, which gives the true performance but is extremely time-consuming. Recently, one-shot NAS substantially reduces the computation cost by training only one supernetwork, a.k.a. supernet, to approximate the performance of every architecture in the search space via weight-sharing. However, the performance estimation can be very inaccurate due to the co-adaption among operations. In this paper, we propose few-shot NAS that uses multiple supernetworks, called sub-supernet, each covering different regions of the search space to alleviate the undesired co-adaption. Since each subsupernet only covers a small search space, compared to one-shot NAS, few-shot NAS improves the accuracy of architecture evaluation with a small increase of evaluation cost. With only up to 7 sub-supernets, few-shot NAS establishes new SoTAs: on ImageNet, it finds models that reach 80.5 top-1 at 600 MB FLOPS and 77.3 top-1 at 230 MFLOPS; on CIFAR10, it reaches 98.72 top-1 without using extra data or transfer learning. In Auto-GAN, few-shot NAS outperforms the previously published results by up to 20\%. Extensive experiments show that few-shot NAS significantly improves various one-shot methods, including 4 gradient-based and 6 search-based methods on 3 different tasks in NASBench-201 and NASBench one-shot-one.