Fast Coverage of N-Dimensional State Space Self-Organizing Map

This paper puts forward a proposal for a new type of Self-Organizing Map which makes fast computations, has topology and number of nodes variable. The map follows a set of 3 basic rules to create groups with receptive fields of approximately the same size regardless of the density of samples. Comparisons of performance with other types of maps were made which used MSE as the metric error and the run time. Two time-series were used as benchmarks. Possible applications of this model include clustering, time-series forecasting and Vector Quantization.