The Relationship of DBSCAN to Matrix Factorization and Spectral Clustering.
Erich Schubert, Sibylle Hess, Katharina J. Morik · TU/e Research Portal · 2018
DBSCAN is a popular approach for density-based clustering. In this short ``work in progress'' paper, we want to present an interpretation of DBSCAN as a matrix factorization problem, which introduces a theoretical connection (but not an equivalence) between DBSCAN and Spectral Clustering (SC). While this does not yield a faster algorithm for DBSCAN, establishing this relationship is a step towards a more unified view of clustering, by identifying further relationships between some of the most popular clustering algorithms.