An Introduction to Neural Architecture Search for Convolutional Networks

George Kyriakides,Konstantinos Margaritis

DOI: https://doi.org/10.48550/arXiv.2005.11074

IF: 5.414

2020-05-22

Machine Learning

Abstract:Neural Architecture Search (NAS) is a research field concerned with utilizing optimization algorithms to design optimal neural network architectures. There are many approaches concerning the architectural search spaces, optimization algorithms, as well as candidate architecture evaluation methods. As the field is growing at a continuously increasing pace, it is difficult for a beginner to discern between major, as well as emerging directions the field has followed. In this work, we provide an introduction to the basic concepts of NAS for convolutional networks, along with the major advances in search spaces, algorithms and evaluation techniques.