Nerve cell synapse circuit and nerve cell circuit

Abstract

The invention discloses a nerve cell synapse circuit and a nerve cell circuit, wherein the nerve cell synapse circuit comprises a charging circuit, a discharging circuit and an MOS (metal oxide semiconductor) capacitor, wherein the MOS capacitor is connected with the charging circuit and the discharging circuit; the charging circuit and the discharging circuit are both formed by a plurality of MOS devices, and are connected into a pulse sequence generated by nerve cells before the synapse and a pulse sequence generated by the nerve cells after the synapse; the charging circuit is constructed to charge the MOS capacitor when the pulse sequence generated by the nerve cells before the synapse is reached earlier than the pulse sequence generated by the nerve cells after the synapse so that the simulation voltage for increasing the synapse weight is output; the discharging circuit is constructed to discharge the MOS capacitor when the pulse sequence generated by the nerve cells before the synapse is reached later than the pulse sequence generated by the nerve cells after the synapse so that the simulation voltage for decreasing the synapse weight is output. The nerve cell synapse circuit and the nerve cell circuit provided by the invention have the advantages that the circuit power consumption can be reduced; the integration degree is improved.

Description

technical field [0001] The invention relates to the technical field of artificial neural networks, in particular to neuron synapse circuits and neuron circuits. Background technique [0002] The human brain has hundreds of millions of neurons, and the number of synapses is even greater. Therefore, power consumption and integration are the two most concerned factors for brain-like neurochips. Brain-like neural chips are considered by scientists to be the most promising technology for the next generation in terms of computing speed, learning mechanism and power consumption. Since the digital storage technology is very mature, its stored weights have high precision, reliable data, mature technology, and standard design, so synapse and neuron circuits are implemented digitally in many schemes. However, with the deepening of artificial neural network research, the shortcomings of traditional digital circuits to realize neural network algorithms are becoming more and more obviou...

AI Technical Summary

Problems solved by technology

However, with the deepening of artificial neural network research, the shortcomings of traditional digital circuits to realize neural network algorithms are becoming more and more obvious.

At this stage, the neuron synaptic circuit required to realize the required multiplication and addition operations and nonlinear transformation is huge in scale, power consumption and volume, and in the analog neural network, it is necessary to combine the synaptic weights between digital and analog The continuous conversion between them requires a large number of D / A and A / D converters, which greatly increases the power consumption of the circuit, making it difficult to meet the needs of development

Images