摘要 |
<p>Artificial neuron integrated circuit provided with a memory (11) for synaptic coefficients, a memory (15) for neuron states, resolution means (11, 13) and learning means (12, 16, 17), which simultaneously operate in parallel on each of the synaptic coefficients in order to determine new synaptic coefficients. The learning means include means (16) for making a learning function act on the input neuron states Vj and on a correction element Si (17) inherent to each output neuron and also include incrementers/ decrementers (12) which determine in parallel the new synaptic coefficients. The learning functions can be formed by AND, exclusive-OR logic gates. The integrated circuit is used in a neural network structure provided with a processing controller (50). Application: neural network structure. <IMAGE></p> |