[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

LOCAL SEARCH TALK: 16th Jan 2001, 2.30pm, Nottingham Univ, UK.




Hi,

The next meeting of the OR Society Local Search Study Group is in 2001 and
details are below. All are welcome.

Happy Xmas and new year to you all BTW.

PS. Please contact the speaker directly for copies of papers etc.

Dr Andrew Tuson MA(Oxon) MSc(Edin) PhD(Edin) GRSC (andrewt@soi.city.ac.uk)

Lecturer, Department of Computing, City University, London, UK.

***********

OPERATIONAL RESEARCH SOCIETY

LOCAL SEARCH STUDY GROUP

Title: Hunting big game in fitness landscapes

Speaker: Colin R Reeves, Coventry University

Venue and Time:	University of Nottingham,
 		School of Computer Science and IT,
		Room C54,
		2.30pm Tuesday 16th January 2001.

Abstract:

Neighbourhood search (NS) is a popular methods for finding good solutions
to combinatorial optimization problems (COPs). Individual neighbourhoods
induce different fitness landscapes, replete with "hills", "valleys",
"ridges" etc., but simple NS only climbs one hill - a local optimum. While
at least one hill will be globally optimal, experience shows that most of
the time we don't find it, and the fitness of the one we do climb may be
quite poor.

It would be helpful if we could have some idea of how many local optima
there really are in a particular fitness landscape. In this talk I shall
demonstrate how this problem can be formulated as an analogy of a
well-know problem in ecology - estimating the numbers of an animal species
in a well-defined area. Some theoretical results will be explained, and
experimental evidence used to assess the plausibility of this approach.
 
Colin Reeves web page is at: http://www.mis.coventry.ac.uk/~colinr/

Directions can be found at: http://www.nottingham.ac.uk/estate/nc.htm