We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Statistics > Methodology

Title: Robust low-rank covariance matrix estimation with a general pattern of missing values

Authors: Alexandre Hippert-Ferrer (1), Mohammed Nabil El Korso (2), Arnaud Breloy (2), Guillaume Ginolhac (3) ((1) L2S, Paris-Saclay University, Paris, France, (2) LEME, Paris Nanterre University, Paris, France, (3) LISTIC, Savoie Mont Blanc University, Annecy, France)
Abstract: This paper tackles the problem of robust covariance matrix estimation when the data is incomplete. Classical statistical estimation methodologies are usually built upon the Gaussian assumption, whereas existing robust estimation ones assume unstructured signal models. The former can be inaccurate in real-world data sets in which heterogeneity causes heavy-tail distributions, while the latter does not profit from the usual low-rank structure of the signal. Taking advantage of both worlds, a covariance matrix estimation procedure is designed on a robust (mixture of scaled Gaussian) low-rank model by leveraging the observed-data likelihood function within an expectation-maximization algorithm. It is also designed to handle general pattern of missing values. The proposed procedure is first validated on simulated data sets. Then, its interest for classification and clustering applications is assessed on two real data sets with missing values, which include multispectral and hyperspectral time series.
Comments: 35 pages, 10 figures
Subjects: Methodology (stat.ME); Signal Processing (eess.SP)
Cite as: arXiv:2107.10505 [stat.ME]
  (or arXiv:2107.10505v2 [stat.ME] for this version)

Submission history

From: Alexandre Hippert-Ferrer [view email]
[v1] Thu, 22 Jul 2021 07:55:25 GMT (1435kb,D)
[v2] Tue, 23 Nov 2021 18:19:24 GMT (1250kb,D)

Link back to: arXiv, form interface, contact.