Author(s)

R. Tsaih

Abstract

Most of promising evolutionary ANNs involve a global search algorithm that is a stochastic one. We have introduced a heuristic learning algorithm for the application problems with non-binary desired outputs. The learning algorithm is named as the SBP learning algorithm, where the softening algorithm is first used to decide network's structure, and the obtained NS is then trained via BP. The design of the SBP learning algorithm does not follow the ideas of genetic inheritance and natural selection; however, during its learning process, NS de- velops in a deterministic way by degrees to a more advanced or mature stage. Thus the SBP learning algorithm is a deterministic evolutionary learning algo- rithm. The arrangement of the SBP learning algorithm is acceptable provided the ability of successfully classifying

Keywords