A neuron hardware device and a method for simulating a spiking neural network using the device

Abstract

The invention discloses a neuron hardware structure and a method for simulating a pulse neural network using the structure. The neuron hardware structure is characterized in that it includes a neural network, and the neural network includes a plurality of neuron layers. The neuron layer comprises a plurality of neurons, and the neuron comprises a synaptic layer, and the synaptic layer comprises a plurality of synapses; the method for simulating the pulse neural network with the above-mentioned neuron hardware structure is characterized in that it comprises the following Steps: 1) Determine the synapse model; 2) Simulate the synapse; 3) Simulate the neuron; 4) Simulate the neuron layer; 5) Simulate the neural network. This neuron hardware structure can reduce the hardware resources occupied by a single neuron node. By using this neuron hardware structure to simulate the spiking neural network, the simulation time is short, the scalability is good, and the hardware resources occupied by the spiking neural network can be reduced, thereby improving the ability of hardware devices to accommodate neurons.

Description

technical field [0001] The invention relates to a large-scale pulse neural network technology, in particular to a method for simulating a pulse neural network with a neuron hardware device. Background technique [0002] The rapid development of neuroscience has accumulated a lot of knowledge about the structure and function of the human brain. Studies have shown that the brain is composed of dense, complex interconnections of neurons, which exhibit many amazing properties, such as pattern recognition and decision-making control. The current understanding of biological neurons is that they transmit information and perform computations through the timing of spikes. The researchers proposed a computational model of spiking neural networks, which simulates behaviors such as information transmission between neurons and signal processing within neurons. At present, there are many fields using computing methods based on spiking neural networks, such as prediction, image processin...

AI Technical Summary

Problems solved by technology

Software modeling and simulating spiking neural networks are relatively easy to implement and have a short development cycle, but software is generally based on von Neumann’s serial execution architecture, so for large-scale neural networks, software simulation requires a lot of simulation time and can be poor scalability

Images