Neural network training method, device and equipment and computer readable storage medium

Abstract

The invention discloses a neural network training method comprising the following steps: randomly initializing a to-be-trained neural network, and obtaining target random parameters of the initializedto-be-trained neural network based on input data; quantizing the target random parameter to obtain a quantized parameter; determining a quantization loss value based on the target random parameter and the quantization parameter; and determining a target neural network based on the quantitative loss value and the initialized to-be-trained neural network. The invention further discloses a neural network training device and equipment and a computer readable storage medium. According to the method, the neural network is trained through the quantized parameters, so that the target neural network obtained through training has high model interpretability, a large number of redundant parameters in the neural network are prevented from being processed by selecting the target random parameters, andthe network calculation amount in the model training process is reduced, so that the neural network can be miniaturized and deployed in small-sized edge equipment.

Description

technical field [0001] The present invention relates to the field of neural networks, in particular to a neural network training method, device, equipment and computer-readable storage medium. Background technique [0002] With the development of artificial intelligence, deep learning has shown great advantages in the fields of image detection and speech recognition. Neural network is an important algorithm for deep learning. However, since there are a large number of redundant parameters in the neural network, the calculation of the neural network model will be huge, so it cannot be directly used in some application scenarios such as embedded devices and other small edge devices. [0003] The above content is only used to assist in understanding the technical solution of the present invention, and does not mean that the above content is admitted as prior art. Contents of the invention [0004] The main purpose of the present invention is to provide a neural network trai...

AI Technical Summary

Problems solved by technology

[0004] The main purpose of the present invention is to provide a neural network training method, device, equipment and computer-readable storage medium, aiming to solve the technical problem of the huge amount of calculation of the existing neural network model

Images