Method for local multi-modal sparse code completion through adaptive similar structure regularization

Abstract

The invention discloses a method for local multi-modal sparse code completion through adaptive similar structure regularization. The method mainly comprises the following steps that firstly, accordingto existing local multi-modal data with the combination of a probability neighbor matrix method, a final target function for local multi-modal sparse code completion through adaptive similar structure regularization is obtained; secondly, according to the obtained target function, a corresponding optimization algorithm is employed to learn a probability neighbor P, a unified sparse code alpha anda multi-modal base vector dictionary D, and accordingly, a final sparse code is obtained. Compared with a common sparse cod solution scheme, the method can reasonably utilize the local multi-modal data, and the sparse code which better meets similarity requirements is generated. Compared with a traditional method, the effect obtained in sparse coding problems is better.

Description

technical field [0001] The invention relates to sparse coding completion, in particular to a method for local multimodal sparse coding completion using adaptive similar structure regularization. Background technique [0002] With the development of the Internet, many Internet applications begin to contain multi-modal data, such as data containing both web page images and related texts. Therefore, the multimodal sparse coding technology for multimodal data has become an important technology. The purpose of sparse coding technology is to represent multimodal data with only a few coding coefficients. [0003] Existing technology is mainly through the use of l 1 The sparse coding regularization method of the penalty term is used to reconstruct the multimodal data to complete the sparse coding and preserve the model correlation through the method of multimodal data mapping function learning. However, in practical situations, generally the data in many modalities are incomplete...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to solve the problems in the prior art. In order to overcome the lack of special attention to local multimodal data in the prior art, the present invention provides a local multimodal sparse coding using adaptive similar structure regularization Complementary method

Images