A MULTI-INTERACTING PERCEPTRON MODEL WITH CONTINUOUS OUTPUTS

We consider learning and generalization of real functions by a multi-i nteracting feed-forward network model with continuous outputs with inv ertible transfer functions. The expansion in different multi-interacti ng orders provides a classification for the functions to be learnt and suggests the learning rules, that reduce to the Hebb-learning rule on ly for the second order, linear perceptron. The over-sophistication pr oblem is straightforwardly overcome by a natural cutoff in the multi-i nteracting synapses: the student is able to learn the architecture of the target rule, that is, the simpler a rule is, the faster the multi- interacting perceptron may learn. Simulation results are in excellent agreement with analytical calculations.