Intelligent Engineering Systems through Artificial Neural Networks Volume 18
60 A Strategy for Minimizing the Processing Time of the aiNet Algorithm in the Construction of Radial Basis Function Neural Networks
Download citation file:
The use of Artificial Intelligence techniques for clustering tasks often requires significant processing time, as it is the case with the aiNET algorithm applied to Radial Basis Function Neural Networks (RBFNN) construction. To minimize processing time, we improved the aiNET algorithm by dividing the internal memory of clonal antibodies into smaller matrices. Interactively, each matrix passing through the clonal suppressing process and the remaining antibodies are concatenated in the next iteration matrix, preserving some past selectivity. In ten RBFNN construction experiments, the internal memory matrix of clonal antibodies was half divided. Consequently, the processing time for the improved aiNET was, on average, 95% smaller than for the standard aiNET. The error rate in misclassification of the resulting RBFNN (11,81% for standard aiNET and 11,68% for improved aiNET) demonstrates that dividing the memory matrix of clonal antibodies is a valid strategy to minimize the processing time for an aiNET algorithm.