Generalising algorithm performance in instance space: A timetabling case study

Kate Smith-Miles, Leonardo Lopes

Research output: Chapter in Book/Report/Conference proceedingConference PaperResearchpeer-review

21 Citations (Scopus)

Abstract

The ability to visualise how algorithm performance varies across the feature space of possible instance, both real and synthetic, is critical to algorithm selection. Generalising algorithm performance, based on learning from a subset of instances, creates a footprint in instance space. This paper shows how self-organising maps can be used to visualise the footprint of algorithm performance, and illustrates the approach using a case study from university course timetabling. The properties of the timetabling instances, viewed from this instance space, are revealing of the differences between the instance generation methods, and the suitability of different algorithms
Original languageEnglish
Title of host publicationLecture Notes in Computer Science [P]
EditorsCarlos A Coello
Place of PublicationGermany
PublisherSpringer
Pages524 - 538
Number of pages15
ISBN (Print)0302-9743
DOIs
Publication statusPublished - 2011
EventInternational Conference on Learning and Intelligent OptimizatioN (LION) 2011 - Rome, Italy
Duration: 17 Jan 201121 Jan 2011
Conference number: 5th
http://www.intelligent-optimization.org/LION5/ (Conference website)
https://link.springer.com/book/10.1007/978-3-642-25566-3 (Proceedings)

Conference

ConferenceInternational Conference on Learning and Intelligent OptimizatioN (LION) 2011
Abbreviated titleLION 5
CountryItaly
CityRome
Period17/01/1121/01/11
OtherProceedings of this conference are published in Lecture Notes in Computer Science
LNCS 6683
Learning and Intelligent Optimization
5th International Conference, LION 5
Rome, Italy, January 17-21, 2011
Selected Papers
Springer
ISSN 0302-9743
e-ISSN 1611-3349
ISBN 978-3-642-25565-6
e-ISBN 978-3-642-25566-3
DOI 10.1007/978-3-642-25566-3
Internet address

Cite this