Greedy Restart Schedules: A Baseline for Dynamic Algorithm Selection on Numerical Black-box Optimization Problems.
Research output: Contribution to book/Conference proceedings/Anthology/Report › Conference contribution › Contributed › peer-review
Contributors
Abstract
In many optimization domains, there are multiple different solvers that contribute to the overall state-of-the-art, each performing better on some, and worse on other types of problem instances. Metaalgorithmic approaches, such as instance-based algorithm selection, configuration and scheduling, aim to close this gap by extracting the most performance possible from a set of (configurable) optimizers. In this context, the best performing individual algorithms are often hand-crafted hybrid heuristics which perform many restarts of fast local optimization approaches. However, data-driven techniques to create optimized restart schedules have not yet been extensively studied.Here, we present a simple scheduling approach that iteratively selects the algorithm performing best on the distribution of unsolved training problems at time of selection, resulting in a problem-independent solver schedule. We demonstrate our approach using well-known optimizers from numerical black-box optimization on the BBOB testbed, bridging much of the gap between single and virtual best solver from the original portfolio across various evaluation protocols. Our greedy restart schedule presents a powerful baseline for more complex dynamic algorithm selection models.
Details
| Original language | English |
|---|---|
| Title of host publication | GECCO '25: Proceedings of the Genetic and Evolutionary Computation Conference |
| Editors | Gabriela Ochoa |
| Pages | 1199 - 1207 |
| Number of pages | 9 |
| ISBN (electronic) | 9798400714658 |
| Publication status | Published - Jul 2025 |
| Peer-reviewed | Yes |
External IDs
| ORCID | /0000-0003-3929-7465/work/188438530 |
|---|---|
| Scopus | 105013080836 |
Keywords
ASJC Scopus subject areas
Keywords
- algorithm scheduling, benchmarking, continuous optimization, dynamic algorithm selection, heuristic optimization, performance analysis