Computational neural networks [CNNs] have been used widely in many application areas in recent years. Most applications use feedforward CNNs and the backpropagation [BP] training algorithm. There are numerous variants of the classical BP algorithm and other training algorithms, but all these training algorithms assume a fixed CNN architecture. They only train weights in the fixed architecture that includes both connectivity and node transfer functions [see also Chapter 8 in this volume]. The problem of designing a near optimal CNN architecture for an application remains unsolved. This is an important issue, because there is strong biological and engineering evidence to support the contention that the function, i.e. the information processing capability of an CNN, is determined by its architecture.
Weitere Kapitel dieses Buchs durch Wischen aufrufen
- Evolving Computational Neural Networks Through Evolutionary Computation
- Springer Berlin Heidelberg
Neuer Inhalt/© ITandMEDIA