A Survey on Neural Architecture Search

PhD Qualifying Examination


Title: "A Survey on Neural Architecture Search"

by

Mr. Shimin DI


Abstract:

Neural Architecture Search (NAS) has shown its power in designing better neural 
network models that can adapt to the given task. It has become a mainstream 
research topic in the machine learning area since the competitive performance 
is achieved on the CIFAR-10 and Penn Treebank benchmarks. However, searching 
proper neural architectures is not easy, which requires plenty of computation 
resources and domain-specific knowledge.

In this survey, we review previous studies on NAS. We first introduce the 
general principles in NAS: search space, search algorithm, and evaluation 
mechanism. Then we focus on the more recent work, One-shot Architecture Search, 
which has been proposed to significantly reduce the search cost in NAS. In the 
end, we summarize the potential opportunities in NAS and discuss future work.


Date:			Monday, 2 November 2020

Time:                  	3:00pm - 5:00pm

Zoom meeting:           https://hkust.zoom.com.cn/j/3060855400

Committee Members:	Prof. Lei Chen (Supervisor)
 			Prof. Dik-Lun Lee (Chairperson)
 			Dr. Qifeng Chen
 			Prof. James Kwok


**** ALL are Welcome ****