PredNAS: A Universal and Sample Efficient Neural Architecture Search Framework
This work addresses the challenge of sample efficiency in NAS for researchers and practitioners, offering a universal framework that reduces computational costs while maintaining performance.
The paper tackles the problem of neural architecture search (NAS) by proposing PredNAS, a framework that uses a differentiable performance predictor and gradient ascent to optimize architectures, achieving state-of-the-art results on NAS benchmarks with less than 100 training samples and competitive performance on large-scale tasks like ImageNet and MSCOCO.
In this paper, we present a general and effective framework for Neural Architecture Search (NAS), named PredNAS. The motivation is that given a differentiable performance estimation function, we can directly optimize the architecture towards higher performance by simple gradient ascent. Specifically, we adopt a neural predictor as the performance predictor. Surprisingly, PredNAS can achieve state-of-the-art performances on NAS benchmarks with only a few training samples (less than 100). To validate the universality of our method, we also apply our method on large-scale tasks and compare our method with RegNet on ImageNet and YOLOX on MSCOCO. The results demonstrate that our PredNAS can explore novel architectures with competitive performances under specific computational complexity constraints.