Information content confirmation method and system based on oscillation neural network

Abstract

The invention provides an information content confirmation method and system based on an oscillation neural network. The method comprises the following steps: encoding an input digital signal into at least one first frequency signal; performing feature abstraction operation on each first frequency signal to generate a corresponding second frequency signal; each second frequency signal is classified, a classification result is output, and the content of the to-be-confirmed information is confirmed based on the classification result. Feature abstraction operation is carried out on the frequency signals, so that the defect that the frequency signals are large in numerical value range and discrete is overcome; the problem that an existing traditional neural network training method cannot be applied to the existing neural network training method is solved, the neural network model adopts frequency signals to confirm information, efficient and accurate oscillation neural network training and reasoning can be achieved, and meanwhile the problem of precision loss caused by actual deployment does not exist.

Description

technical field [0001] The invention belongs to the field of artificial intelligence, and in particular relates to a method and system for confirming information content based on an oscillation neural network. Background technique [0002] With the continuous development of information technology and the growing demand of people, people have higher and higher requirements for the timeliness of information. Artificial neural networks are widely used in the field of computer vision, and have obtained great achievements in image classification, target recognition and video monitoring. more mature applications. Oscillatory neural network (ONN) is a special neural network that uses frequency / phase to store information. Compared with traditional neural networks, it greatly reduces the amount of calculation and required memory, and has its unique advantages. [0003] However, due to the large and discrete characteristics of the frequency signal, the existing traditional neural net...

AI Technical Summary

Problems solved by technology

[0003] However, due to the large and discrete characteristics of the frequency signal, the existing traditional neural network training methods cannot be applied to it. The existing spiking neural network (SNN) is also a special neural network, which fully considers the time of the signal. feature, which delivers information driven by event pulses

In the prior art, the training method adopted by SNN is not very mature. After processing the signal in an approximate way, the gradient backpropagation method is used to train the network. The learning efficiency is low, and even the convergence cannot be achieved, and the SNN signal is quite different from the ONN signal, so the pulse The learning method of the neural network cannot be applied to the oscillatory neural network, and there are many deficiencies

Images