An improved n-dimensional self-organizing neural network simulator

A description is given of enhancements made to a self-organizing neural network simulator that is used in the identification of lithofacies pattern features. The original simulator was developed for n-dimensional data and used spare matrix techniques to reduce processor memory requirements. The simulator obtained good results but required relatively long computing times. For this reason, the simulator was studied with the goal of improving overall performance. Those program areas requiring the majority of processing time were analyzed and redesigned. As a result, the performance of the simulator was dramatically improved. An overview of both the original simulator and the modified version is presented. Modifications which were made are presented along with data from performance tests