Heterogeneous transfer learning method based on optimal subspace learning

Abstract

The invention discloses a heterogeneous transfer learning method based on an optimal subspace. The method comprises the following steps: extracting source domain and target domain data for heterogeneous transfer learning from a database; a model is constructed based on typical correlation analysis and a multi-classification loss function, so that a target function can find out the most correlatedfeature mapping subspace of the two domains, and features in the feature space can have strong classification capability; equivalent transformation is carried out on the target function to simplify the model; processing the simplified model, dividing an optimization problem of the whole model into three sub-problems by using an alternating direction multiplier method, and solving to obtain an updated solution of a parameter variable in the three sub-problems; and finally, carrying out iterative updating on the whole problem to converge to the optimal. The method well combines two characteristics of knowledge migration capability and model classification capability, can be well applied to a difficult heterogeneous migration learning scene, and has very strong classification learning and data labeling capability in the application of an actual scene.

Description

technical field [0001] The invention relates to the fields of heterogeneous transfer learning and semi-supervised transfer learning, in particular to a heterogeneous transfer learning method based on optimal subspace learning. Background technique [0002] In recent years, artificial intelligence technology has developed rapidly and played a strong role in various field application scenarios. The core of artificial intelligence technology is the learning and prediction of its algorithms. However, with the increase in the diversity and complexity of practical application scenarios. The shortcomings of traditional artificial intelligence algorithms are becoming more and more prominent. This is mainly due to the fact that traditional artificial intelligence algorithms assume that the learned data obeys the same distribution, and only by obeying such an assumption can they show good results. However, in practical applications, due to various reasons such as different times an...

AI Technical Summary

Problems solved by technology

However, in practical applications, due to various reasons such as different times and different regions, the data often do not completely obey the same distribution, which requires transfer learning technology to apply the learned knowledge from the source domain to the target domain.

A class of difficult and valuable practical problems is that not only the distribution of the source domain and the target domain are different, but also the characteristics of the data are also different. This is the heterogeneous transfer learning problem.

At present, many researchers have paid attention to this problem, but the effective solutions to this problem are very limited.

Images