Analysis of multi-step algorithms for cognitive maps learning

This article is devoted to the analysis of multi-step algorithms for cognitive maps learning. Cognitive maps and multi-step supervised learning based on a gradient method and unsupervised one based on the non-linear Hebbian algorithm were described. Comparative analysis of these methods to one-step algorithms, from the point of view of the speed of convergence of a learning algorithm and the influence on the work of the decision systems was performed. Simulation results were done on prepared software tool ISEMK. Obtained results show that implementation of the multi-step technique gives certain possibilities to get quicker values of target relations values and improve the operation of the learned system.