Modified Hebbian learning rule for single layer learning

The author introduce a new learning method based on the supervised Hebbian learning of encoded associations. The supervised Hebbian learning formula is characterized by rather quick convergence. This feature can be helpful especially for large networks and for a large amount of input data. In addition, initial weights in the proposed rule do not have to be assumed as is the case in the Hebbian rule near zero. The rule was implemented for the case of three training point clusters and compared with other rules suitable for providing the classification of the training set data. The method is suitable for training of suitable-layer networks