A reconfigurable self-learning spiking neural network processor

Abstract

The invention provides a reconfigurable self-learning pulse neural network processor, which includes a processing unit array composed of a plurality of processing units and channels in four directions of east, south, west and north; the processing unit includes an external routing module, a sorting module, pulse queue module, controller module, search module, memory module, client module, server module, internal routing module, multiple exclusive computing resources and multiple borrowable computing resources; the processor adopts pulse generation time and source neural The pulse packet transmission signal composed of element ID; the mode of computing resource execution calculation is divided into reasoning mode and learning mode, and the reconfigurable circuit is used to update the target neuron membrane potential, synaptic weight related variables or synaptic weight; computing resource Including adaptive clock-driven and event-driven computing mechanism modules, according to the update time interval, adaptively change the calculation method of the calculation unit to perform update calculations.

Description

technical field [0001] The invention relates to the technical field of pulse neural network, in particular to a reconfigurable self-learning pulse neural network processor. Background technique [0002] Neural network is a mathematical model abstracted according to the characteristics of biological brain nervous system, which can be used to solve practical problems such as object recognition. Its development process can be divided into three generations: Perceptron, Artificial Neural Network (ANN), Pulse Neural Network (Spiking Neural Network, SNN). Early perceptrons could only solve linear problems, but were powerless to nonlinear problems (such as XOR problems); ANN added a nonlinear activation function on the basis of perceptrons to enable it to solve nonlinear problems. In recent years, with the improvement of computer computing power, ANN has a better performance in solving some complex problems, but with the increase of the complexity of the problem, a deeper network ...

AI Technical Summary

Problems solved by technology

Due to the relatively large network structure, this design requires high transmission bandwidth of the system, resulting in reduced system performance

[0006] 3. In order to support the online learning of SNN, the existing design needs to design an additional learning circuit, which greatly increases the overhead of hardware resources

Therefore, for different pulse input situations, different update methods should be used in order to make the power consumption of the system as low as possible, but the current design does not take this into consideration.

Images