Skip to content
Longterm Wiki
Back

Advances in neural architecture search

web

Credibility Rating

5/5
Gold(5)

Gold standard. Rigorous peer review, high editorial standards, and strong institutional reputation.

Rating inherited from publication venue: Oxford Academic

Relevant to AI safety discussions around automated capability improvement and intelligence explosion hypotheses; NAS represents a concrete instantiation of machines improving machine learning systems, though this paper focuses on technical advances rather than safety implications.

Metadata

Importance: 35/100journal articlereference

Summary

This academic survey reviews progress in Neural Architecture Search (NAS), covering automated methods for designing neural network architectures. It examines search strategies, performance estimation techniques, and applications across various domains, highlighting how NAS enables automated discovery of architectures that rival or surpass hand-designed models.

Key Points

  • NAS automates the design of neural network architectures, reducing reliance on human expertise and enabling systematic exploration of architectural space.
  • Key NAS strategies include reinforcement learning, evolutionary algorithms, gradient-based methods, and Bayesian optimization.
  • Performance estimation techniques like weight sharing and proxy tasks have dramatically reduced the computational cost of NAS.
  • NAS is increasingly applied to specialized domains including computer vision, NLP, and hardware-aware deployment.
  • Automated architecture discovery is a form of recursive capability improvement, raising questions about the pace and controllability of AI advancement.

Cited by 2 pages

Cached Content Preview

HTTP 200Fetched Apr 9, 202698 KB
Advances in neural architecture search | National Science Review | Oxford Academic
 
 

 

 

 

 

 
 
 
 

 

 

 

 

 

 

 
 

 

 

 

 

 

 

 
 
 
 

 

 

 

 
 
 
 

 Jan
 FEB
 Mar
 

 
 

 
 05
 
 

 
 

 2025
 2026
 2027
 

 
 
 

 

 

 
 
success

 
fail

 
 
 
 
 
 
 
 
 
 
 

 

 
 
 
 
 
 
 
 
 

 

 About this capture
 

 

 

 

 

 

 
COLLECTED BY

 

 

 
 
Collection: Save Page Now Outlinks

 

 

 

 

 
TIMESTAMPS

 

 

 

 

 

 

The Wayback Machine - https://web.archive.org/web/20260205104248/https://academic.oup.com/nsr/article/11/8/nwae282/7740455

 

 

 
 
 

 
 

 

 

 

 
 
 
 
 
 
 

 

 
 
 
 
 
 
 
 
 

 

 

 

 
 
 
 Skip to Main Content

 

 

 

 

 

 

 

 

 
 

 
Advertisement

 

 
 

 

 

 

 
 
 

 

 

 

 Journals
 

 

 Books
 

 

 
 

 

 
 

 
 Search Menu
 
 

 
 

 
 AI Discovery Assistant
 
 
 

 

 
 
 
 

 

 
 
 
 

 
 

 
 Menu
 
 

 

 

 
 
 
 

 

 
 
 
 

 

 

 

 
 Sign in through your institution
 

 
 

 
 Remember my
institution
 

 

 
 

 

 

 

 

 

 

 

 

 
 

 

 
 

 
 Navbar Search Filter
 
 

National Science ReviewThis issueComputer Science
 
Medicine and HealthScience and MathematicsBooksJournalsOxford Academic 
 

 
 Mobile Enter search term
 
 

 
 Search
 

 

 

 

 

 
 

 
 

 
 Issues
 
 
 

 

 Subject

 
 

 

 Chemistry
 

 

 Computer Science
 

 

 Earth Sciences
 

 

 Information Science
 

 

 Life Sciences
 

 

 Materials Science
 

 

 Mathematics
 

 

 Physics
 

 

 Science Policy
 

 

 
 

 

 
 More Content
 
 
 
 
 

 
 Advance Access
 
 
 

 

 
 Special Topics
 
 
 

 

 

 

 
 Submit
 
 
 
 
 

 
 Author Guidelines
 
 
 

 

 
 Submission Site
 
 
 

 

 
 Open Access Options
 
 
 

 

 
 Self-Archiving Policy
 
 
 

 

 
 Reasons to submit
 
 
 

 

 

 

 
 Alerts
 
 
 

 

 
 About
 
 
 
 
 

 
 About National Science Review
 
 
 

 

 
 Editorial Board
 
 
 

 

 
 Advertising and Corporate Services
 
 
 

 

 
 Journals Career Network
 
 
 

 

 
 Policies
 
 
 

 

 
 Dispatch Dates
 
 
 

 

 

 

 Journals on Oxford Academic
 

 

 Books on Oxford Academic
 

 
 

 

 

 
 
 

 

 
 
 

 

 

 

 
 

 
 Issues
 
 
 

 

 
 Subject
 
 
 

 
 
 

 
 
 All Subject
 
 
 Expand 
 
 
 Expand 
 
 
 

 
 

 
 

 Chemistry

 

 

 
 

 

 Computer Science

 

 

 
 

 

 Earth Sciences

 

 

 
 

 

 Information Science

 

 

 
 

 

 Life Sciences

 

 

 
 

 

 Materials Science

 

 

 
 

 

 Mathematics

 

 

 
 

 

 Physics

 

 

 
 

 

 Science Policy

 

 

 
 

 
 

 

 Browse all content
 Browse content in 
 

 
 

 

 
 More Content
 
 
 
 
 

 
 Advance Access
 
 
 

 

 
 Special Topics
 
 
 

 

 

 

 
 Submit
 
 
 
 
 

 
 Author Guidelines
 
 
 

 

 
 Submission Site
 
 
 

 

 
 Open Access Options
 
 
 

 

 
 Self-Archiving Policy
 
 
 

 

 
 Reasons to submit
 
 
 

 

 

 

 
 Alerts
 
 
 

 

 
 About
 
 
 
 
 

 
 About National Science Review
 
 
 

 

 
 Editorial Board
 
 
 

 

 
 Advertising and

... (truncated, 98 KB total)
Resource ID: d1a3f270ea185ba1 | Stable ID: sid_1EoqQJQiAv