Fuzzy clustering method and device based on anchor graph

Abstract

The invention discloses a fuzzy clustering method and device based on an anchor graph. The method comprises the following steps: selecting a plurality of anchor points from an original data matrix ofan image to form an anchor graph matrix; multiplying the anchor graph matrix by an inverse matrix of the diagonal matrix, and performing singular value decomposition on the multiplied matrix to obtainspectrum embedded data; establishing a relationship between a global fuzzy membership matrix and an anchor graph matrix; establishing a fuzzy clustering target function based on the anchor graph according to the relationship between the fuzzy membership matrix and the anchor graph matrix, wherein a regular term is set in the target function; iteratively solving the fuzzy clustering objective function based on the anchor graph until the fuzzy clustering objective function based on the anchor graph converges; and discretizing the spectrum embedded data by adopting a converged fuzzy clustering target function based on the anchor graph to obtain a clustering result. According to the invention, the relationship between the anchor graph and the fuzzy membership degree can be better obtained, the fuzzy membership degrees of the data points and all clustering centers are balanced, and the more reasonable fuzzy membership degree is obtained.

Description

technical field [0001] The present application relates to the technical field of image clustering, in particular to an anchor graph-based fuzzy clustering method and device. Background technique [0002] In recent years, Fast Spectral Clustering based on Anchor Graph (FSCAG) has received more and more attention due to its excellent performance. The FSCAG algorithm mainly consists of two steps, which are constructing the similarity graph based on anchor points and discretizing the spectral embedding data by using K-means. The FSCAG algorithm uses anchor points to represent the original data points to construct a similarity graph, which greatly reduces the computational complexity and effectively solves the problem that the spectral clustering algorithm is difficult to obtain better clustering results in large-scale data. Since the adjacency matrix in the FSCAG algorithm is composed of similarity graphs of data samples and their anchor points, its spectral embedding results a...

AI Technical Summary

Problems solved by technology

[0004] This application provides a fuzzy clustering method and device based on anchor graphs, which solves the lack of robustness in the abnormal environment of calculating t

Images