| Author: | Arnold L. Patton |
| Advisor: | William F. Punch III, Erik D. Goodman |
| Email: | pattona@cps.msu.edu; http://www.cps.msu.edu/~pattona79 |
Evolutionary Computation (EC) operators typically act in a strictly stochastic manner which is wholy dependent on the shape of the problem landscape being searched and the effects of random initialization. Thus, there is little the user of such systems can do to modify the progress of the search or the balance between exploration and exploitation. This is especially problematic in production systems where the user often wishes to impose time constraints on the search. In this poster, we outline an approach published in the Proceedings of the Congress on Evolutionary Computation [Patton 99], which allows for the progress of the search to be scheduled according to the measure of the variance loss during each generation.
Return to Workshop Listing