Population diversity as an objective in EMO

by Amit


Population Diversity is a key aspect in Evolutionary Algorithms ( Image source: http://thistimethisspace.com/2008/04/22/rising-population/)

When you send a paper to a conference or a journal, a definitive positive outcome is the reviews that you get- all the more so when you have sent in your work to a reputed conference/journal. Whether your paper gets accepted or rejected is ofcourse the next thing, and often the important thing. However, the reviews can improve your work in either case.

In my recent work, which I shall definitely report ( :-) )after I have got the official notification, one of the reviewers was kind enough to point to me some works in the literature who have used a measure of population diversity to obtain better solutions to Optimization problems.

Here are the works that will give you an idea about the approach used in Evolutionary Multi-Objective Optimization (EMO):

Multi-objective diversity maintenance

@inproceedings{Snijders2006, author = {Snijders, Paul and de Jong, Edwin D. and de Boer, Bart and Weissing, Franjo},         title = {Multi-objective diversity maintenance}, booktitle = {GECCO ’06: Proceedings of the 8th annual conference on Genetic and evolutionary computation}, year = {2006}, isbn = {1-59593-186-4}, pages = {1429–1430}, location = {Seattle, Washington, USA}, doi = {http://doi.acm.org/10.1145/1143997.1144229}, publisher = {ACM}, address = {New York,NY, USA}, }

Reducing bloat and promoting diversity using multi-objective methods
@conference{de2001reducing,
title={{Reducing bloat and promoting diversity using multi-objective methods}},
author={De Jong, E.D. and Watson, R.A. and Pollack, J.B.},
booktitle={Proceedings of the Genetic and Evolutionary Computation Conference},
pages={11–18},
year={2001}
}
Genetic Diversity as an Objective in Multi-Objective Evolutionary Algorithms

@article{Toffolo2003, author = {Andrea Toffolo and Ernesto Benini}, title = {Genetic Diversity as an Objective in Multi-Objective Evolutionary Algorithms}, journal = {Evolutionary Computation}, volume = {11}, number = {2}, year = {2003}, pages = {151-167}, bibsource = {DBLP, http://dblp.uni-trier.de} }

The basic idea in these workis to use population diversity as a second objective (for a single-objective optimization problem), thus making it a Multi-Objective Optimization (MOO) problem. Authors have found that doing the above enables maintain population diversity (maximizing the second objective) and hence find optimal solutions, which otherwise may not have been found.

You may also be interested to read my earlier post Measure of population diversity in Genetic Algorithms

Advertisements