Circuit for pre-charging the input vector components in a free neuron circuit during the recognition phase

摘要

In a neural network comprised of a plurality of neuron circuits being either in an engaged or a free state, there is disclosed a pre-charge circuit placed in each neuron circuit that allows to load the components of an input vector (A) only in a determined free neuron circuit during the recognition phase as the components of the prototype vector (B) attached to this neuron circuit. The pre-charge circuit consists of a weight memory (251) controlled by a memory control signal (RS) and of a circuit generating said memory control signal which is adapted to identify the said determined free neuron circuit. During the recognition phase, the memory control signal is thus set active only for the said determined free neuron circuit. When the neural network is structured as a chain of neuron circuits, said determined free neuron circuit is the first free in the chain. The input data bus (DATA-BUS) transporting the input vector components is connected to the weight memory of all neuron circuits. The data outputted therefrom are available on an output data bus (RAM-BUS). The pre-charge circuit may further include an address counter (252) to properly address the weight memory and a register (253) to latch the data that are outputted on the output data bus. After this neuron circuit has been engaged, the contents of its weight memory cannot be modified any longer. To perform the pre-charge of the input vector during the recognition phase makes the engagement process more efficient and significantly reduces the learning time of the input vector.