P. Fenton and P. Walsh (Ireland)
Genetic Algorihtms, Seeding, Scheduling.
The generic and robust nature of Genetic Algorithms (GA) has often led to its coupling with other techniques, in order to provide it with added capability in dealing with complex problems. In this work we present an examination of one such coupling for a permutation with repetition representation based GA applied to a job shop scheduling problem (JSP). We seed a GA with chromosomes created by a number of heuristics and dispatch rules and from the output of previous runs. Our motivation stems from the lack of analysis of the performance of seeding for this representation in this domain. We seek to determine whether by improving the average fitness of the initial population, we can improve the performance of the GA. Mutation rates are also examined in an attempt to increase the diversity of the GA population. We conclude that an optimum level of seeding exists, above which increased mutation rates cannot do enough to offset the problems arising due to lack of diversity.
Important Links:
Go Back
