Neuron weight information processing method and system

Abstract

The invention relates to a neuron weight information processing method and system. The method comprises the steps that the neuron output information of a front end neuron and a weight index corresponding to the neuron output information are received; weight index information correspondence relation is read according to the weight index, and weight information is acquired, wherein the weight index information correspondence relation is correspondence between the weight index and the weight information; and the input information of the front end neuron is acquired according to the weight information and the neuron output information. Neurons no longer directly transmit the weight information, but transmit the index of the weight information. The amount of information transmission between networks is saved. The setting of the weight information can be flexibly changed. The information processing capacity of a neuron network is improved.

Description

technical field [0001] The invention relates to the technical field of artificial neural networks, in particular to a neuron weight information processing method and system. Background technique [0002] Most of today's artificial neural network research is still implemented in von Neumann computer software and high-performance GPGPU (General Purpose Graphic Processing Units) platform. The hardware overhead, energy consumption and information of the whole process The processing speed is not optimistic. For this reason, the field of neuromorphic computing has developed rapidly in recent years, that is, using hardware circuits to directly construct neural networks to simulate the functions of the brain, trying to achieve a computing platform that is massively parallel, low-energy, and capable of supporting complex pattern learning. [0003] However, in the traditional neuromorphic system, when binding the connection and weight information between neurons in the neural network...

AI Technical Summary

Problems solved by technology

[0003] However, in the traditional neuromorphic system, when binding the connection and weight information between neurons in the neural network, the connection between each neuron needs to transmit the corresponding weight information, which occupies more hardware Processing resources and hardware have great constraints on network performance, and weight information reduces the processing power of neurons and limits the flexibility of neurons. It is not suitable for general-purpose neural network hardware frameworks.

Images