Method for storing computer neural consisting of neurons for use in medical research, involves generating identity signals representative of degree of identity between identity vector and reference vector of neurons

摘要

The method involves grouping neurons into aggregates, where each aggregate includes neurons in a same layer. An inhibition signal is generated for activating neurons in each aggregate, and identity signals representative of a degree of identity between an identity vector consisting of input signals of the neurons and a reference vector of the neurons are generated. The identity signals are integrated and compared with a threshold, and the reference vector of the neurons is updated by convergence towards an input vector that is maintained until the neurons are active. An independent claim is also included for a computer neural consisting of elementary circuits identified as neurons.