Show simple item record

dc.contributor.authorSmythe, Jared Michael Dean
dc.description.abstractIn this paper a new probability-based multi-valued Particle Swarm Optimization algorithm is developed for problems with nominal variables. The algorithm more explicitly contextualizes probability updates in terms of roulette wheel probabilities. It is first applied to three forest planning problems, and its performance is compared to the performance of a forest planning algorithm, as well as to results from other algorithms in other papers. The new algorithm outperforms the others except in a single case, where a customized forest planning algorithm obtains superior results. Finally, the new algorithm is compared to three other probability-based Particle Swarm Optimization algorithms via analysis of their probability updates. Three intuitive but fully justified requirements for generally effective probability updates are presented, and through single iteration and balance convergence tests it is determined that the new algorithm violates these requirements the least.
dc.subjectMulti-valued particle swarm optimization
dc.subjectDiscrete particle swarm optimization
dc.subjectForest planning
dc.subjectRaindrop optimization
dc.subjectNominal variables
dc.subjectCombinatorial optimization
dc.subjectProbability optimization
dc.titleRoulette Wheel Particle Swarm Optimization
dc.description.departmentArtificial Intelligence Center
dc.description.majorArtificial Intelligence
dc.description.advisorWalter D. Potter
dc.description.committeeWalter D. Potter
dc.description.committeeMichael Covington
dc.description.committeePete Bettinger

Files in this item


There are no files associated with this item.

This item appears in the following Collection(s)

Show simple item record