Methods of gene adaptation in case of genetically evolved artificial neural networks

Artificial neural networks are successfully applied not only for solving problems which cannot be described by mathematical formulae, but also for optimization of existing solutions. Interesting possibility for building proper neural networks can be application of evolutionary algorithms. There are three methods of preparing neural networks discussed in this article: weight-based, neuron-based and path-based adaptation. What is more, the article discusses behavior of the genetic algorithm depending on different methods of removing weights in the network.