New PLSCAN algorithm offers improved multiscale density-based clustering

Researchers have introduced PLSCAN, a novel multiscale density-based clustering algorithm designed for exploratory data analysis. PLSCAN addresses the challenge of hyperparameter selection in existing density-based methods like DBSCAN and HDBSCAN by employing a persistence-based cluster selection procedure. This approach identifies stable clusters across various scales, offering improved performance and stability compared to HDBSCAN* on real-world datasets, as indicated by higher median ARI and better resampling stability. Additionally, PLSCAN demonstrates competitive run times against k-Means++ on lower-dimensional data. AI

RANK_REASON The item is a research paper detailing a new algorithm for data analysis. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• DBSCAN
• HDBSCAN*
• K-means++
• PLSCAN
• Seigneur de Lormois

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →