Algorithm Selection
Overview
Algorithm selection studies how to choose among available algorithms when their performance varies across problem instances, settings, or resource limits. Common approaches use meta-learning and per-instance features, learned representations such as graphs or landscape images, algorithm portfolios, and online methods that switch strategies as new evidence arrives. The goal is to improve on a single fixed choice while accounting for the cost of measuring instance features, running candidates, and deploying the selector. A central challenge is whether selection learned on benchmarks transfers reliably to unseen real-world problems and whether its end-to-end gains persist after those costs are included.
Latest papers
50 papers5 papers in the last four weeks, down 17% on the four weeks before. 0.0% of all new papers.