Deep convolution neural network training method and device

Abstract

The present invention relates to the field of deep learning techniques, in particular to a deep convolution neural network training method and a device. The deep convolution neural network training method and the device comprise the steps of a, pretraining the DCNN on a large scale data set, and pruning the DCNN; b, performing the migration learning on the pruned DCNN; c, performing the model compression and the pruning on the migrated DCNN with the small-scale target data set, In the process of migrating learning of large-scale source data set to small-scale target data set, the model compression and the pruning are performed on the DCNN by the migration learning method and the advantages of model compression technology, so as to improve the migration learning ability to reduce the risk of overfitting and the deployment difficulty on the small-scale target data set and improve the prediction ability of the model on the target data set.

Description

technical field [0001] The invention relates to the technical field of deep learning, in particular to a deep convolutional neural network training method and device. Background technique [0002] In recent years, with the rapid development of Internet and computer technology, Deep Convolutional Neural Network (DCNN) has achieved breakthrough success in challenging subjects such as image classification and audio recognition. However, the model structure of DCNN is huge and complex, which requires large-scale data to optimize and train model parameters. However, many practical problems in real life are usually only supported by small-scale data, and it is difficult to obtain a high-performance DCNN by directly using the small-scale training data of the target task. A widely used strategy is transfer learning, which is an effective technique for modeling small-scale target datasets in the field of deep learning research. A large number of studies have shown that DCNNs traine...

AI Technical Summary

Problems solved by technology

However, from the perspective of model compression, DCNN’s pruning compression strategy [Hinton et al., 2015; Han et al., 2015; Han et al., 2016] is mainly for the operation of the same large-scale source data set, and does not involve Transfer Learning for Small Target Datasets

Therefore, such methods still have the problem of how to build a high-performance DCNN for small-scale target datasets.

Images