Self-adaptation of mutation rates in non-elitist populations
- Submitting institution
-
The University of Birmingham
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 53936241
- Type
- E - Conference contribution
- DOI
-
10.1007/978-3-319-45823-6_75
- Title of conference / published proceedings
- PPSN 2016 : Parallel Problem Solving from Nature – PPSN XIV
- First page
- 803
- Volume
- 9921
- Issue
- -
- ISSN
- 0302-9743
- Open access status
- Technical exception
- Month of publication
- August
- Year of publication
- 2016
- URL
-
-
- Supplementary information
-
-
- 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
-
1
- Research group(s)
-
-
- Citation count
- 24
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- This paper proves for the first time that self-adaptation can be an effective parameter-control mechanism in discrete evolutionary algorithms (EAs). This result is significant, because parameter tuning accounts for 10-25% of the cost of industrial applications of EAs. Empirical research suggested that self-adaptation, where parameter settings evolve with the population, could be beneficial. This paper provided for the first time a mathematical analysis of when self adaptation of mutation-rates can lead to exponential speed-ups compared with fixed mutation rates. Ideas from the paper have been implemented in the Nevergrad toolbox published by Facebook Research.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -