Unsupervised feature selection method and system based on multi-label learning

Abstract

The invention provides an unsupervised feature selection method and system based on multi-label learning, and the method comprises the steps: carrying out feature extraction of each obtained data sample, obtaining a feature data set, learning a binary multi-label matrix and a feature selection matrix for the feature data set, and constructing an unsupervised feature selection target function basedon multi-label learning; solving an unsupervised feature selection target function based on multi-label learning by adopting a discrete optimization method based on an augmented Lagrange multiplier method to obtain a feature selection matrix; and sorting the feature selection matrix to determine selected target features. Learning multiple tags for semantic guidance and executing feature selection, and applying binary constraints in spectrum embedding to obtain multiple tags to guide a final feature selection process. In addition, a dynamic sample similarity graph is constructed in a self-adaptive mode to capture a data structure, so that the multi-label discrimination capability is enhanced.

Description

technical field [0001] The present disclosure relates to the technical field of feature selection, in particular to an unsupervised feature selection method and system based on multi-label learning. Background technique [0002] With the rapid development of information technology, high-dimensional data emerges in different research fields, such as multimedia computing, data mining, pattern recognition and machine learning, etc. On the one hand, high-dimensional data can provide richer information. On the other hand, it also brings a challenging curse of dimensionality problem. High-dimensional data usually contains noise or outliers, so directly using such high-dimensional data will often have a bad impact on subsequent learning tasks, and even reduce the performance of the method. In order to solve this problem, dimensionality reduction technology is proposed, which includes two different processing methods: (1) feature selection; (2) feature extraction. [0003] Featur...

AI Technical Summary

Problems solved by technology

[0004] Although good performance has been achieved, there are still some problems that need to be further resolved: (1) Existing unsupervised feature selection methods either have no label guidance, or use a single label to guide the process of feature selection; the former makes the selected Feature semantics are missing, the latter will cause information loss

(2) The quality of the graph created by the existing graph-based feature selection method is not high, and the graph is usually constructed directly on the original data through the Gaussian kernel, and the graph remains fixed throughout the model learning process

In addition, the graph creation process and feature selection process are separated into two independent processes, which will also make the method produce suboptimal results.

Images