Deep migration learning system and method based on entropy minimization

Abstract

The invention provides a deep migration learning method based on entropy minimization, and the method comprises the steps: S1) dividing a source field and a target field according to different migration learning tasks, constructing a migration learning network, and initializing hyper-parameters of the migration learning network; S2) inputting respective data samples of the source domain and the target domain into a transfer learning network and carrying out forward propagation to obtain a network prediction label; training the whole network by using a stochastic gradient descent method according to the proposed loss function, completing updating of network parameters by using back propagation, and stopping training until the model converges or reaches the maximum number of iterations; S3)storing the network model and a training result; according to the target domain label prediction method based on the entropy minimization technology. By maximizing the distribution diversity of each batch of data prediction categories in the target domain, a common solution result occurring when only the entropy minimization technology is used is avoided, and the reliability of transfer learning is guaranteed.

Description

technical field [0001] The present invention relates to a learning system, specifically a deep transfer learning method, Background technique [0002] Supervised training of machine learning on a large amount of labeled data can achieve good performance and results. However, large labeled data sets are limited in number and application fields, and manual labeling of sufficient training data often requires high costs. Therefore, when faced with a target task where labeled data is scarce, how to use the existing labeled data in the source domain that is related to the target domain but obeys a different probability distribution to build an effective learner has a strong practical demand. [0003] Aiming at this problem, transfer learning method is usually used to solve it, that is, to train a discriminator to adjust the parameters of the transfer learning network, so that under the transfer learning network after parameter adjustment, the distribution offset between the data i...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a deep transfer learning method based on entropy minimization to solve the ordinary solution problem that occurs when only entropy minimization technology is used

Images