TANGO: Clustering with Typicality-Aware Nonlocal Mode-Seeking and Graph-Cut Optimization

2024-08-19

arXiv.org

AI Summary

Density-based clustering suffers from over-reliance on local structures and insufficient global characterization, leading to spurious peak detection and high sensitivity to threshold parameters. To address these issues, this paper proposes a typicality-aware non-local pattern optimization framework. Its key contributions are: (1) a novel global typicality-based dependency recalibration mechanism that eliminates reliance on local density estimation; (2) integration of path connectivity measurement with graph-cut optimization to achieve centerless, adaptive cluster partitioning; and (3) an efficient typicality estimation algorithm coupled with a non-local density propagation strategy. Evaluated on 16 real-world and multiple synthetic benchmarks, the method consistently outperforms mainstream algorithms—including DBSCAN and MeanShift—in clustering accuracy and robustness. It exhibits strong noise resilience and is fully parameter-free, requiring no manual hyperparameter tuning.

Technology Category

• Machine Learning: Clustering
• Search and Optimization: Local Search
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Web Mining and Content Analysis: Normalization, clustering, classification, and summarization of Web text
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Density-based clustering methods by mode-seeking usually achieve clustering by using local density estimation to mine structural information, such as local dependencies from lower density points to higher neighbors. However, they often rely too heavily on emph{local} structures and neglect emph{global} characteristics, which can lead to significant errors in peak selection and dependency establishment. Although introducing more hyperparameters that revise dependencies can help mitigate this issue, tuning them is challenging and even impossible on real-world datasets. In this paper, we propose a new algorithm (TANGO) to establish local dependencies by exploiting a global-view emph{typicality} of points, which is obtained by mining further the density distributions and initial dependencies. TANGO then obtains sub-clusters with the help of the adjusted dependencies, and characterizes the similarity between sub-clusters by incorporating path-based connectivity. It achieves final clustering by employing graph-cut on sub-clusters, thus avoiding the challenging selection of cluster centers. Moreover, this paper provides theoretical analysis and an efficient method for the calculation of typicality. Experimental results on several synthetic and $16$ real-world datasets demonstrate the effectiveness and superiority of TANGO.

Problem

• Improves mode-seeking by using global typicality instead of local thresholds
• Introduces TANGO for clustering with typicality-aware nonlocal mode-seeking
• Enhances clustering via graph-cut optimization and path-based similarity

Innovation

• Uses typicality for global-view mode detection
• Employs graph-cut with path-based similarity
• Combines nonlocal mode-seeking and optimization