|
Abstract:
|
We theoretically explore some of the properties of evolutionary algorithms (EAs). We discover that under certain conditions, it is more advantageous to utilize a restarting procedure for the evolutionary algorithm than to continue to allow the algorithm to run due to an exponentially increasing time required for transitions between optima. In the cases that restarting is indicated, we discover that an exhaustive algorithm based on a given evolutionary algorithm may be able to outperform the evolutionary algorithm on which it is based. We demonstrate the application of this algorithm to the BQP problem on test problems found in the literature, recovering the best performance reported in the literature.
|
