A Neuromorphic Processor Based on Parameter Quantization Sharing

Abstract

The invention relates to the technical field of neuromorphic hardware, in particular to a neuromorphic processor for parameter quantization sharing. The processor of the present invention introduces a parameter quantization sharing structure in the existing neuromorphic processor architecture, specifically through the quantization parameter control module, which is used to obtain information from the outside of the neuromorphic processor (external storage, external Host computer, etc.) reads the quantization parameter and writes it into the quantization parameter storage module, and directly reads the current synapse type from outside the neuromorphic processor (external storage, external host computer, etc.) during the neuromorphic processor running phase, and According to the read synapse type, the quantized parameter corresponding to the synapse type is read from the quantized parameter storage module to configure the neuron computing module. Therefore, the storage space of the processor is effectively saved, the processor area is greatly reduced, the computing efficiency of the processor is greatly improved, and the computing power consumption is reduced.

Description

technical field [0001] The invention relates to the technical field of neuromorphic hardware, in particular to a neuromorphic processor for parameter quantization sharing. Background technique [0002] The term Neuromorphic was first proposed by American scientist and engineer Carver Mead in the form of "neuromorphic processors" in the late 1980s. Neuromorphic hardware (neuromorphic hardware) is to efficiently abstract and simulate biological nervous systems through memristors, threshold switches, or analog, digital, analog / digital hybrid VLSI hardware systems, in order to achieve similar On the basis of the information processing ability of the biological nervous system, it achieves the characteristics of low power consumption and high adaptability. [0003] Existing neuromorphic hardware needs to pre-store a number of parameters equal to the number of neurons in the neuromorphic network to be calculated in order to complete the calculation operation of the network. Gener...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the problem that current neuromorphic processors require a large amount of storage space to store network parameters, the present invention proposes a neuromorphic processor based on parameter quantization sharing, which introduces a parameter quantization sharing structure into the existing neuromorphic processor architecture , so as to effectively save the storage space of the processor, greatly reduce the area of ​​the processor, greatly improve the computing efficiency of the processor, and reduce the computing power consumption

Images