Configurable solvers: Tailoring general methods to specific applications
Applying constraint-based problem solving methods in a new domain often requires considerable work. In this talk I will examine the state of the art in constraint-based problem solving techniques and the difficulties involved in selecting and tuning an algorithm to solve a problem. Most constraint-based solvers have many algorithmic variations, and it can make a very significant difference exactly which algorithm is used and how the problem is encoded. I will describe promising new approaches in which generic algorithms are automatically configured for specific applications.
Unable to display preview. Download preview PDF.
- 1.J. Allen and S. Minton. Selecting the right heuristic algorithm: Runtime performance predictors. In Proceedings of the Canadian AI Conference, 1996.Google Scholar
- 2.D.J. Cook and R.Craig Varnell. Maximizing the benefits of parallel search using machine learning. In Proceedings AAAI-97, 1997.Google Scholar
- 3.S. Minton. An analytic learning system for specializing heuristics. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence, 1993.Google Scholar
- 4.S. Minton. Automatically configuring constraint satisfaction programs: A case study (in press. Constraints, 1(1), 1996.Google Scholar
- 5.D.R. Smith. KIDS: A knowledge-based software development system. In M.R. Lowry and R.D. McCartney, editors, Automating Software Design. AAAI Press, 1991.Google Scholar