Neural network, information processing method thereof, and information processing system thereof

Abstract

Disclosed are a neural network, an information processing method thereof, and an information processing system thereof. The neural network includes N neuron layers connected one by one, and except forthe first neuron layer, each neuron of each neuron layer includes m dendritic units and one hippocampal unit. Each dendritic unit includes a resistance gradient device, each hippocampal unit includesa resistance mutation device, and the m dendritic units can be respectively set with different threshold voltages or currents; and the neurons of the n-th neuron layer are respectively connected withthe m dendritic units of the neurons of the n+1-th neuron layer. N is an integer greater than or equal to 3, m is an integer greater than 1, and n is an integer greater than or equal to 1 and less than N. The neural network can filter most non-key information through the dendritic units, thus, the information processing amount of the neural network is reduced, the power consumption of the neuralnetwork operation is reduced, and the processing efficiency of key information by the neural network is increased.

Description

technical field [0001] Embodiments of the present disclosure relate to a neural network, an information processing method thereof, and an information processing system. Background technique [0002] With the rapid development of information technology, neural networks have achieved remarkable results in many fields such as speech and image recognition. However, with the continuous strengthening of neural network capabilities, the demand for computing resources is also increasing, and the von Neumann architecture, which separates data processing units and data storage units, has lagged far behind information-emerging systems. Speed ​​has become a bottleneck restricting its development. Therefore, inspired by the biological neural network, the distributed information processing system (that is, the artificial neural network) that combines the data processing unit and the data storage unit has become one of the most effective ways to solve the data processing bottleneck of the...

AI Technical Summary

Problems solved by technology

However, with the continuous strengthening of neural network capabilities, the demand for computing resources is also increasing, and the von Neumann architecture, which separates data processing units and data storage units, has lagged far behind information-emerging systems. speed, which has become a bottleneck restricting its development

Images