Abstract
We present a comparison of step size adaptation methods for evolution strategies, covering recent developments in the field. Following recent work by Hansen et al. we formulate a concise list of performance criteria: a) fast convergence of the mean, b) near-optimal fixed point of the normalized step size dynamics, and c) invariance to adding constant dimensions of the objective function. Our results show that algorithms violating these principles tend to underestimate the step size or are unreliable when the function does not fit to the algorithm's tuned hyperparameters. In contrast, we find that cumulative step size adaptation (CSA) and twopoint adaptation (TPA) provide reliable estimates of the optimal step size. We further find that removing the evolution path of CSA still leads to a reliable algorithm without the computational requirements of CSA.
Original language | English |
---|---|
Title of host publication | Proceedings of the 14th ACM/SIGEVO Conference on Foundations of Genetic Algorithms |
Number of pages | 10 |
Publisher | Association for Computing Machinery |
Publication date | 2017 |
Pages | 139-148 |
ISBN (Electronic) | 978-1-4503-4651-1 |
DOIs | |
Publication status | Published - 2017 |
Event | 14th ACM/SIGEVO Conference on Foundations of Genetic Algorithms - København, Denmark Duration: 12 Jan 2017 → 15 Jan 2017 Conference number: 14 |
Conference
Conference | 14th ACM/SIGEVO Conference on Foundations of Genetic Algorithms |
---|---|
Number | 14 |
Country/Territory | Denmark |
City | København |
Period | 12/01/2017 → 15/01/2017 |
Keywords
- Comparison
- Evolution strategies
- Step size adaptation