Yao Shu, Shaofeng Cai, Zhongxiang Dai, Beng Chin Ooi, Bryan Kian Hsiang Low
Recent years have witnessed a surging interest in Neural Architecture Search (NAS). Various algorithms have been proposed to improve the search efficiency and effectiveness of NAS, i.e., to reduce the search cost and improve the generalization performance of the selected architectures, respectively. However, the search efficiency of these algorithms is severely limited by the need for model training during the search process. To overcome this limitation, we propose a novel NAS algorithm called NAS at Initialization (NASI) that exploits the capability of a Neural Tangent Kernel in being able to characterize the converged performance of candidate architectures at initialization, hence allowing model training to be completely avoided to boost the search efficiency. Besides the improved search efficiency, NASI also achieves competitive search effectiveness on various datasets like CIFAR-10/100 and ImageNet. Further, NASI is shown to be label- and data-agnostic under mild conditions, which guarantees the transferability of architectures selected by our NASI over different datasets.
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Neural Architecture Search | NATS-Bench Topology, CIFAR-10 | Test Accuracy | 93.55 | NASI (Shu et al., 2021) |
| Neural Architecture Search | NATS-Bench Topology, CIFAR-100 | Test Accuracy | 71.2 | NASI (Shu et al., 2021) |
| Neural Architecture Search | NATS-Bench Topology, ImageNet16-120 | Test Accuracy | 44.84 | NASI (Shu et al., 2021) |
| AutoML | NATS-Bench Topology, CIFAR-10 | Test Accuracy | 93.55 | NASI (Shu et al., 2021) |
| AutoML | NATS-Bench Topology, CIFAR-100 | Test Accuracy | 71.2 | NASI (Shu et al., 2021) |
| AutoML | NATS-Bench Topology, ImageNet16-120 | Test Accuracy | 44.84 | NASI (Shu et al., 2021) |