An Improved R4-rule for Evolutionary Learning of NN-MLP

W. Du and Q. Zhao (Japan)


The R4 -rule, nearest neighbor based multilayer perceptron (NN-MLP), pattern recognition, learning vector quantization, and neural networks


To design the nearest-neighbor-based multilayer perceptron (NN-MLP) efficiently, we have proposed a non-genetic evolutionary algorithm called the R4 -rule. Experimental results obtained so far show that the R4 rule can produce the smallest or nearly smallest networks with high generalization ability by iteratively performing four basic operations: recognition, remembrance, reduction, and review. To use this algorithm, however, we must specify several parameters properly, and this is often not easy. In this paper, we try to improve the usability of the R4 -rule by reducing the number of parameters. The efficiency of the improved algorithm is verified through experiments with several public databases.

Important Links:

Go Back