BlockSwap: Fisher-guided Block Substitution for Network Compression on a Budget
- Submitting institution
-
University of Edinburgh
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 156893313
- Type
- E - Conference contribution
- DOI
-
-
- Title of conference / published proceedings
- Proceedings to the International Conference on Learning Representations 2020
- First page
- 1
- Volume
- -
- Issue
- -
- ISSN
- -
- Open access status
- -
- Month of publication
- April
- Year of publication
- 2020
- URL
-
-
- Supplementary information
-
https://iclr.cc/virtual_2020/poster_SklkDkSFPB.html
- Request cross-referral to
- -
- Output has been delayed by COVID-19
- No
- COVID-19 affected output statement
- -
- Forensic science
- No
- Criminology
- No
- Interdisciplinary
- No
- Number of additional authors
-
4
- Research group(s)
-
B - Data Science and Artificial Intelligence
- Citation count
- -
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- In this work, we have shown that is possible to search for neural architectures in a matter of seconds, instead of days. The paper was accepted at ICLR, which is a top-tier machine learning conference (26.5% acceptance rate of 2594 submissions). It attracted interest from a colleague within Edinburgh’s Usher Institute, which has lead to an on-going collaboration and a follow up paper (under review), the code for which has been starred over 100 times on github. A YouTube video about this follow-up work has been made by a popular vlogger, garnering thousands of views in a matter of days.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -
