A dimensionality reduction method combining graph optimization and projection learning

Abstract

The invention relates to a dimensionality reduction method combining graph optimization and projection learning, and belongs to the field of pattern recognition and machine learning. Firstly, graph optimization and projection matrix learning are integrated into a unified framework, the graph structure can be learned in a self-adaptive mode in the dimensionality reduction process, and the graph structure can well describe the geometric structure of data. In the framework, through adopting l21-norm based distance measurement function, negative effects caused by abnormal values or data changes are reduced, thereby improving the robustness of the algorithm. In addition, local constraints are introduced to ensure that neighbor area samples are selected as much as possible for reconstruction, and therefore it is ensured that the algorithm can well keep local structure information of high-dimensional data. Finally, the invention provides an effective iteration updating algorithm to solve themodel. According to a large number of experiments, experimental results verify that the method has good performance, is superior to an existing related method, and is suitable for classification and clustering tasks of high-dimensional data.

Description

technical field [0001] The invention relates to the fields of computer vision technology, pattern recognition and machine learning, in particular to a graph dimension reduction method. Background technique [0002] In many fields of machine learning and computer vision, high-dimensional data usually contains a large number of redundant and noisy features, which will lead to the "curse of dimensionality" problem and reduce the effectiveness of existing algorithms. Therefore, how to use dimensionality reduction (dimensionality reduction) technology to extract the most useful low-dimensional representation from the original high-dimensional data has become a key issue. [0003] In the past few decades, a series of dimensionality reduction (dimensionality reduction) algorithms have been proposed, and these methods can be divided into two categories: linear and nonlinear. The linear dimension reduction method maps the original high-dimensional data to a low-dimensional subspace ...

AI Technical Summary

Problems solved by technology

Therefore, the method is still very sensitive to outliers or changes in the data, and in addition, its optimization process is relatively complex and time-consuming

Images