Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Clustering Based on Density Propagation and Subcluster Merging

Feiping Nie, Yitao Song, Jingjing Xue, Rong Wang, Xuelong Li

TL;DR

The DPSM method is proposed, a density-based node clustering approach that automatically determines the number of clusters and can be applied in both data space and graph space, and the concept of spectral clustering is extended from individual nodes to small clusters, while introducing the CluCut measure to guide cluster merging.

Abstract

We propose the DPSM method, a density-based node clustering approach that automatically determines the number of clusters and can be applied in both data space and graph space. Unlike traditional density-based clustering methods, which necessitate calculating the distance between any two nodes, our proposed technique determines density through a propagation process, thereby making it suitable for a graph space. In DPSM, nodes are partitioned into small clusters based on propagated density. The partitioning technique has been proved to be sound and complete. We then extend the concept of spectral clustering from individual nodes to these small clusters, while introducing the CluCut measure to guide cluster merging. This measure is modified in various ways to account for cluster properties, thus provides guidance on when to terminate the merging process. Various experiments have validated the effectiveness of DOSM and the accuracy of these conclusions.

Clustering Based on Density Propagation and Subcluster Merging

TL;DR

The DPSM method is proposed, a density-based node clustering approach that automatically determines the number of clusters and can be applied in both data space and graph space, and the concept of spectral clustering is extended from individual nodes to small clusters, while introducing the CluCut measure to guide cluster merging.

Abstract

We propose the DPSM method, a density-based node clustering approach that automatically determines the number of clusters and can be applied in both data space and graph space. Unlike traditional density-based clustering methods, which necessitate calculating the distance between any two nodes, our proposed technique determines density through a propagation process, thereby making it suitable for a graph space. In DPSM, nodes are partitioned into small clusters based on propagated density. The partitioning technique has been proved to be sound and complete. We then extend the concept of spectral clustering from individual nodes to these small clusters, while introducing the CluCut measure to guide cluster merging. This measure is modified in various ways to account for cluster properties, thus provides guidance on when to terminate the merging process. Various experiments have validated the effectiveness of DOSM and the accuracy of these conclusions.

Paper Structure

This paper contains 19 sections, 20 equations, 7 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Propagation Procedures for Graph Learning
  4. Density Peak Clustering
  5. Spectral Clustering
  6. Proposed Method, DPSM
  7. Evaluating Nodes by Density Propagation
  8. Partitioning Nodes With Propagated Densities
  9. Merging Clusters via Inter- and Intra-cluster Relationships
  10. Similarities and Differences With Other Algorithms
  11. Compared With DPC
  12. Compared With Spectral Clustering
  13. Experimental Results and Analysis
  14. Datasets and Metrics
  15. Clustering Experiments With Unknown Number of Classes
  16. ...and 4 more sections

Figures (7)

  • Figure 1: Schematic diagram of density propagation
  • Figure 2: The Probability Density Distribution of Density After Specific Iterations
  • Figure 3: Schematic Diagrams of the Density-based Node Partitioning Approaches.
  • Figure 4: Comparison Between Implementations of Methods
  • Figure 5: Experiments on the Necessity of Propagated Density (the USPS dataset) The images in the first and second rows illustrate the densities and categories of nodes, whereas those in the third and fourth rows depict the initial and final outcomes of clustering.
  • ...and 2 more figures

Theorems & Definitions (4)

  • proof
  • proof
  • proof
  • proof