An Evolutionary Algorithms to Training Neural Networks for a Two-Spiral Problem

Jinn-Moon Yang,Cheng-Yan Kao
2000-07-10
Abstract:An evolutionary algorithm is introduced to train neural networks for a two-spiral problem. The proposed approach automatically achieves the balance of the solution quality and convergence speed by integrating multiple mutations, family competition, and adaptive rules. Following the description of implementation details, our approach is applied to a two-spiral problem. Experimental results indicate that the proposed approach is able to stably solve this problem and is very competitive with the comparative evolutionary algorithms.
What problem does this paper attempt to address?