A Modified General Regression Neural Network (MGRNN) with new, efficient training algorithms as a robust ‘black box’-tool for data analysis

• Published in: Neural networks Vol. 14; no. 8; pp. 1023 - 1034
• Main Authors: Tomandl, Dirk, Schober, Andreas
• Format: Journal Article
• Language: English
• Published: Oxford Elsevier Ltd 01.10.2001; Elsevier Science
• Subjects: Algorithms; Applied sciences; Artificial intelligence; Classification; Computer science; control theory; systems; Connectionism. Neural networks; Exact sciences and technology; General Regression Neural Network; GRNN; Intertwined spirals; Kernel regression; Mackey-Glass; Neural Networks (Computer); P roben1; PROBEN1; Regression; Regression Analysis; Statistics as Topic - methods; Time-series prediction; User-Computer Interface; Time-series prediction; Mackey–Glass; Kernel regression; Classification; Regression; General Regression Neural Network; GRNN; P roben1; Intertwined spirals; Data analysis; Network architecture; Neural network; Fast algorithm; Feedforward; Learning algorithm
• ISSN: 0893-6080; 1879-2782
• DOI: 10.1016/S0893-6080(01)00051-X

A Modified General Regression Neural Network (MGRNN) is presented as an easy-to-use ‘black box’-tool to feed in available data and obtain a reasonable regression surface. The MGRNN is based on the General Regression Neural Network by D. Specht [Specht, D. (1991). A General Regression Neural Network. IEEE Transactions on Neural Networks, 2(6), 568–576], therefore, the network's architecture and weights are determined. The kernel width of each training sample is trained by two supervised training algorithms. These fast and reliable algorithms require four user-definable parameters, but are robust against changes of the parameters. Its generalization ability was tested with different benchmarks: intertwined spirals, Mackey–Glass time series and P roben1. The MGRNN provides two additional features: (1) it is trainable with arbitrary data as long as a suitable metric exists. Particularly, it is unnecessary to force the data structure to vectors of equal length; (2) it is able to compute the gradient of the regression surface as long as the gradient of the metric is definable and defined. The MGRNN solves common practical problems of common feed-forward networks.