作者： A. R. Webb , D. S. Broomhead , D. Lowe
DOI:
关键词:
摘要: Abstract : Take a feed-forward layered network (such as multilayer perceptron or radial basis function network) which is to operate pattern classifier. The may have several hidden layers, many nodes required and any desired nonlinearities on the units. transfer functions of output should be linear. If trained (using appropriate problem) minimise sum squared error over all outputs patterns such that weights minimum norm, then values for subsequent input will constant. keywords: Radar, Great Britain.