Fusion spiking neural network information processing method and fusion spiking neural network

Abstract

The invention relates to a fusion spiking neural network information processing method and a fusion spiking neural network, and the method comprises the steps of receiving a configuration instruction, and reading the types of spiking neural networks needing to be configured in the configuration instruction, the types of the spiking neural networks comprising a general spiking neural network and a specific spiking neural network; dividing an address space of a memory according to the type of the spiking neural network, and writing configuration information into a register; and reading the configuration information, selecting the corresponding control unit to call the pulse calculation unit to perform pulse calculation based on the configuration information, and storing a pulse calculation result in the memory. Compared with the prior art, the invention has the advantages that the defects that a general spiking neural network is large in connection storage area and a specific spiking neural network is low in universality are overcome, meanwhile, the number of transistors of a chip can be reduced, the size of the chip is reduced, and the efficiency of the chip is improved under the condition that the number of the transistors is the same.

Description

technical field [0001] The invention relates to the field of neural network chips, in particular to a fusion pulse neural network information processing method and a fusion pulse neural network. Background technique [0002] Spiking neural network (SNN) has attracted the attention of academia and industry in recent years due to its low power consumption and characteristics closer to the human brain. In a spiking neural network, the axon is the unit that receives pulses, and the neuron is the unit that sends pulses. A neuron is connected to multiple axons through dendrites, and the connection point between dendrites and axons is called a synapse. After the axon receives the pulse, all dendrites that have synaptic connections with this axon will receive the pulse, which in turn affects the downstream neurons of the dendrite. A neuron sums spikes from multiple axons and sends a spike downstream if the value exceeds a threshold. The pulse neural network propagates a 1-bit puls...

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to provide a fusion pulse neural network information processing method and a fusion pulse neural network, which realizes the multiplexing of the general pulse neural network and the pulse calculation unit and the storage unit of the specific pulse neural network, and the multiplexing degree is high. It overcomes the disadvantages of large connection storage area of ​​general-purpose spiking neural network and low versatility of specific spiking neural network, and at the same time can reduce the number of transistors on the chip, reduce the cost, and improve the performance of the chip with the same number of transistors

Images