Funded by the European Union. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Education and Culture Executive Agency (EACEA). Neither the European Union nor EACEA can be held responsible for them.


Neural Architecture Search: A Survey

Users: 1 - Average Rating: 4.00

The article is about automated neural architecture search (NAS) methods. Automated search methods have arisen attention because manually adjusting of parameters by human experts, which is the current practice, is a time-consuming and prone to errors process. This article reviews the existing work in this field of research, dividing the material according to three dimensions, which are the three key steps of NAS methods: search space (defining which architectures can be represented), search strategy (detailing how to explore the search space, often exponentially large or even unbounded), and performance estimation strategy (the process of estimating the performance of each tested architecture).

Scientific Paper

Data Analytics, Machine Learning

Target Group:

Not available

Cite as:
Elsken, T. and Metzen, J. H. and Hutter F., Neural architecture search: A survey, The Journal of Machine Learning Research 20.1 (2019): 1997-2017.

Author of the review:
Giulia Cademartori
University of Genoa


You have to login to leave a comment. If you are not registered click here

Stefano Zampini

Interesting reading if you want to explore how to automatize the design of Neural Network architecture. This is in fact a review of existing works in this field and can be a good starting point to approach this subject!