On the effects of node duplication and connection-oriented constructivism in neural XCSF
Proceedings of the 10th annual conference companion on Genetic and evolutionary computation
Simulated evolution under multiple criteria conditions revisited
WCCI'08 Proceedings of the 2008 IEEE world conference on Computational intelligence: research frontiers
Instruction selection for ARM/Thumb processors based on a multi-objective ant algorithm
CSR'06 Proceedings of the First international computer science conference on Theory and Applications
| Hi-index | 0.00 |
Self-adaptation of the mutation strengths is a powerful mechanism in evolution strategies (ES), but it can fail. As a consequence premature convergence or ending up in a local optimum in multi-modal fitness landscapes can occur. In this article a new approach controlling the process of self-adaptation is proposed. This approach combines the old ideas of gene deletion and gene duplication with the self-adaptation mechanism of the ES. Gene deletion and gene duplication is used to vary the number of independent mutation strengths. In order to demonstrate the practicability of the new approach several multi-modal test functions are used. Methods from statistical design of experiments and regression tree methods are applied to improve the performance of a specific heuristic-problem combination.