Algebraic Approach and Optimal Physical Clusterization in Interpolation Problems of Artificial Intelligence

摘要

Perceptron-type interpolation systems of artificial intelligence are considered. A concept of optimal physical clusterization allows us to divide a second layer of hidden units into the compact sets of units (clusters). Then, an algebraic approach developed for pattern recognition systems may be extended to other systems. To solve the problems of process forecasting, a data sample should be transformed into a single-moment sample.