Arbeitspapier

Adaptive Nonparametric Clustering

This paper presents a new approach to non-parametric cluster analysis called Adaptive Weights Clustering (AWC). The idea is to identify the clustering structure by checking at different points and for dierent scales on departure from local homogeneity. The proposed procedure describes the clustering structure in terms of weights wij each of them measures the degree of local inhomogeneity for two neighbor local clusters using statistical tests of "no gap" between them. The procedure starts from very local scale, then the parameter of locality grows by some factor at each step. The method is fully adaptive and does not require to specify the number of clusters or their structure. The clustering results are not sensitive to noise and outliers, the procedure is able to recover dierent clusters with sharp edges or manifold structure. The method is scalable and computationally feasible. An intensive numerical study shows a state-of-the-art performance of the method in various articial examples and applications to text data. Our theoretical study states optimal sensitivity of AWC to local inhomogeneity.

Sprache
Englisch

Erschienen in
Series: IRTG 1792 Discussion Paper ; No. 2018-018

Klassifikation
Wirtschaft
Mathematical and Quantitative Methods: General
Thema
adaptive weights
clustering
gap coecient
manifold clustering

Ereignis
Geistige Schöpfung
(wer)
Efimov, Kirill
Adamyan, Larisa
Spokoiny, Vladimir
Ereignis
Veröffentlichung
(wer)
Humboldt-Universität zu Berlin, International Research Training Group 1792 "High Dimensional Nonstationary Time Series"
(wo)
Berlin
(wann)
2018

Handle
Letzte Aktualisierung
20.09.2024, 08:22 MESZ

Datenpartner

Dieses Objekt wird bereitgestellt von:
ZBW - Deutsche Zentralbibliothek für Wirtschaftswissenschaften - Leibniz-Informationszentrum Wirtschaft. Bei Fragen zum Objekt wenden Sie sich bitte an den Datenpartner.

Objekttyp

  • Arbeitspapier

Beteiligte

  • Efimov, Kirill
  • Adamyan, Larisa
  • Spokoiny, Vladimir
  • Humboldt-Universität zu Berlin, International Research Training Group 1792 "High Dimensional Nonstationary Time Series"

Entstanden

  • 2018

Ähnliche Objekte (12)