Stronger NAS with Weaker Predictors

Neural Architecture Search (NAS) often trains and evaluates a large number of architectures. Recent predictor-based NAS approaches attempt to address such heavy computation costs with two key steps: sampling some architecture-performance pairs and fitting a proxy accuracy predictor. Given limited samples, these predictors, however, are far from accurate to locate top architectures due to the difficulty of fitting the huge search space. This paper reflects on a simple yet crucial question: if our final goal is to find the best architecture, do we really need to model the whole space well?. We propose a paradigm shift from fitting the whole architecture space using one strong predictor, to progressively fitting a search path towards the high-performance sub-space through a set of weaker predictors. As a key property of the proposed weak predictors, their probabilities of sampling better architectures keep increasing. Hence we only sample a few well-performed architectures guided by the previously learned predictor and estimate a new better weak predictor. This embarrassingly easy framework produces coarse-to-fine iteration to refine the ranking of sampling space gradually. Extensive experiments demonstrate that our method costs fewer samples to find top-performance architectures on NAS-Bench-101 and NAS-Bench-201, as well as achieves the state-of-the-art ImageNet performance on the NASNet search space. In particular, compared to state-of-the-art (SOTA) predictor-based NAS methods, WeakNAS outperforms all of them with notable margins, e.g., requiring at least 7.5x less samples to find global optimal on NAS-Bench-101; and WeakNAS can also absorb them for further performance boost. We further strike the new SOTA result of 81.3% in the ImageNet MobileNet Search Space. The code is available at this https URL.