Title: NAS-Bench-1Shot1: Benchmarking and Dissecting One-shot Neural Architecture Search
Abstract: One-shot neural architecture search (NAS) has played a crucial role in making NAS methods computationally feasible in practice. Nevertheless, there is still a lack of understanding on how these weight-sharing algorithms exactly work due to the many factors controlling the dynamics of the process. In our paper we introduce a general framework for comparing and analyzing the anytime performance of one-shot Neural Architecture Search methods by leveraging the large-scale tabular benchmark NAS-Bench-101. My talk will first introduce various Neural Architecture Search methods and then explain One-shot Neural Architecture Search in detail.
