A general framework for functional networks

摘要: In this paper, we introduce functional networks as a generalization and extension of the standard neural in sense that every problem can be solved by network also formulated network. But, more importantly, give examples problems cannot using but naturally networks. Functional are defined collection connected units on set nodes. A unit or neuron connects input nodes to output The values calculated from given functions one several arguments. main differences with (a) multivariate different (in which case, no weights necessary, because they subsume functions) (b) outputs coupled, is, coincident. This mathematical model parallels printed circuit boards electronic components, thus giving an intuitive interpretation interesting natural additional application. existence common leads equations whose solution lead substantial simplification initial topology involved. Two types (the one-layer serial networks) discussed detail. For networks, very simple algorithm is given. systems obtained. methods illustrated applications. © 2000 John Wiley & Sons, Inc.