作者： Devert Wicker
DOI:
关键词: Pattern recognition (psychology) 、 Pattern recognition 、 Linear discriminant analysis 、 Perceptron 、 Artificial neural network 、 Correlation 、 Computer science 、 Cluster analysis 、 Discriminant function analysis 、 Artificial intelligence 、 Selection (genetic algorithm) 、 Machine learning
摘要: Abstract : This introduces methods of applying the Network Discriminant Function (NDF) to training and automatic architecture selection feed-forward multiple layer perceptrons (MLPs) used for Pattern Recognition (PR). Knowledge NDF may be impact MLP choices. The gives a new insight into performance in that it reflects how well is clustering data. If an intrinsic capability which makes perform well, then should correlate with other measures such as Sum Squared Error. Architecture choices technique need selected aid ability MLP. major goals research are establish useful criteria compare Cascade-Correlation's error correlation. NDF-Cascade (NDFC) learning proposed, tested evaluated.