A TREE-LIKE PERCEPTRON AND A METHOD FOR PARALLEL DISTRIBUTED TRAINING OF SUCH PERCEPTRONS

摘要

A neural network (20) has an input layer (26) comprising a plurality of neurons (28), a hidden layer (24) comprising a plurality of neurons (22), each neuron being either inhibitory or excitatory, and an output layer (30) comprising at least one neuron (32). Each synapse (34) of a first set of synapses connects a neuron in the input layer to a neuron in the hidden layer. Each synapse (36) of a second set of synapses connects a neuron in the hidden layer to a neuron in the output layer. With a plurality of output neurons, a plurality of parallel output paths is thus created. Each neuron in the hidden layer receives a reinforcement signal from its corresponding neuron in the output layer.