Self-adaptive evolutionary constructive and pruning algorithm (SAECPA) is a new structural hearing algorithm, which be planned for proposing of artificial neural networks (ANNs). SAECPA begins with set of ANN and it?s a simplest formation one hidden neuron is linked towards single input node, in that network intersect and network transmutation which increases the network inhabitants, then using cluster pruning (CP) and survival selection (SS) to prune the network.??As a manifestation of the method, SAECPA is concerned to the forecasting problem - the Mackey-Glass time series. Here user defined constraints intersect probability (pc) and transmutation probability (pm) are considered as input, but it may well be developed self-adaptive to enlarge the unknown neurons and links further proficiently.
展开▼
