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

Abstract

The embodiment of the invention discloses a neural network training method and device, equipment and a computer readable storage medium. The method comprises the following steps: converting parameters of a preset network layer in a to-be-trained model from a first data type to a second data type to obtain a prediction model; performing forward propagation on the obtained training sample in the prediction model until the training sample is transmitted to an output layer of which the parameter type is the first data type, and obtaining a prediction value of the first data type; according to the difference between the predicted value of the first data type and the true value of the training sample, performing parameter updating on the to-be-trained model to obtain an updated model; continuously updating the parameters of the updated model until a training end condition is met, and generating a target model; the target model is used for determining a plurality of prediction labels corresponding to a plurality of label dimensions for an input prediction sample. According to the invention, the efficiency of neural network training can be improved.

Description

technical field [0001] The present disclosure relates to computer vision technology, and in particular to a neural network training method, device, equipment and computer-readable storage medium. Background technique [0002] Image classification is an important issue in the field of computer vision and is widely used in different fields of production and life, such as retail product recognition, industrial defect recognition, vehicle type recognition in monitoring scenarios, municipal monitoring, etc. With the development of machine learning and deep learning technology, deep convolutional neural network is used to extract image features, and the extracted image features are used for classification, and excellent classification results have been achieved. Prior to this, deep convolutional neural networks needed to learn massive amounts of training data to obtain network models available in actual scenarios. However, in related technologies, the training process of the neur...

AI Technical Summary

Problems solved by technology

However, in related technologies, the training process of neural network usually takes a long time, so the efficiency of neural network training is low

Images