Deep learning model training method, device and equipment and storage medium

Abstract

The method is applicable to the technical field of deep learning, and provides a deep learning model training method, device and equipment and a storage medium. The method comprises the following steps: according to the received training image set, using a preset feature extraction algorithm and a pre-constructed untrained deep learning model to extractcorresponding target features, performing local training on a hidden layer of the deep learning model through a target feature set formed by the extracted target features; after local training is completed, selecting a target image set from thetraining image set; performing classification training on the full connection layer of the deep learning model. The deep learning model is trained through the deep learning model, so that the number of samples for training the deep learning model is reduced, the trained deep learning model better conforms to the human brain visual cortex characteristics, the noise resistance and the displacement resistance of the deep learning model are improved, and the training speed and the training effect of the deep learning model are improved.

Description

technical field [0001] The invention belongs to the technical field of deep learning, and in particular relates to a training method, device, equipment and storage medium of a deep learning model. Background technique [0002] Existing deep learning models map a large amount of data into a high-dimensional space through continuous nonlinear transformation by building a deep network to obtain abstract feature vectors, and then complete the task of classifying or regressing data. For deep networks, depth and data volume are the basis for correct expression learning. Generally, the deeper the network, the stronger its nonlinear expression ability, and the larger the data volume, it can fit the data more effectively, which is beneficial to Networks perform classification or regression tasks in high-dimensional spaces. Recently, the accuracy of image recognition on the dataset ImageNet using a deep network structure of more than 150 layers can exceed the accuracy of human eye re...

AI Technical Summary

Problems solved by technology

[0003] The object of the present invention is to provide a training method, device, equipment and storage medium of a deep learning model, aiming at solving the problem that the prior art cannot provide an effective The training method of the deep learning model leads to the problems of large number of samples for training the deep learning model, slow model training, and poor training effect

Images