Abstract
Despite both the commercial and academic success of optimization technology and specifically constraint programming, using the technology still requires significant expertise. For non-trivial applications the quality of a system still has much to do with the quality of the person that implemented it. We investigate algorithm control techniques aimed at achieving strong scheduling performance using off-the-shelf algorithms without requiring significant human expertise. Rather than building knowledge-intensive models relating algorithm performance to problem features, we base the control decisions on the evolution of solution quality over time. Such an approach is crucial to our goal of the reduction of expertise.
This work has received support from SFI under Grant 00/PI.1/C075, the Embark Initiative of IRCSET under Grant PD2002/21 and ILOG, SA.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Carchrae, T., Beck, J.: Low-knowledge algorithm control. In: Proceedings of the Nineteenth National Conference on Artificial Intelligence, AAAI 2004 (2004)
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2004 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Carchrae, T. (2004). Long-Term Learning for Algorithm Control. In: Wallace, M. (eds) Principles and Practice of Constraint Programming – CP 2004. CP 2004. Lecture Notes in Computer Science, vol 3258. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-540-30201-8_71
Download citation
DOI: https://doi.org/10.1007/978-3-540-30201-8_71
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-23241-4
Online ISBN: 978-3-540-30201-8
eBook Packages: Springer Book Archive