Fast estimation of the median covariation matrix with application to online robust principal components analysis

DOI10.1007/S11749-016-0519-XzbMATH Open1373.62124arXiv1504.02852OpenAlexW2963891697MaRDI QIDQ89458FDOQ89458

Hervé Cardot, Antoine Godichon-Baggioni, Antoine Godichon-Baggioni

Publication date: 8 December 2016

Published in: Test, TEST (Search for Journal in Brave)

Abstract: The geometric median covariation matrix is a robust multivariate indicator of dispersion which can be extended without any difficulty to functional data. We define estimators, based on recursive algorithms, that can be simply updated at each new observation and are able to deal rapidly with large samples of high dimensional data without being obliged to store all the data in memory. Asymptotic convergence properties of the recursive algorithms are studied under weak conditions. The computation of the principal components can also be performed online and this approach can be useful for online outlier detection. A simulation study clearly shows that this robust indicator is a competitive alternative to minimum covariance determinant when the dimension of the data is small and robust principal components analysis based on projection pursuit and spherical projections for high dimension data. An illustration on a large sample and high dimensional dataset consisting of individual TV audiences measured at a minute scale over a period of 24 hours confirms the interest of considering the robust principal components analysis based on the median covariation matrix. All studied algorithms are available in the R package Gmedian on CRAN.

Full work available at URL: https://arxiv.org/abs/1504.02852

zbMATH Keywords

functional data\(L_1\)-mediangeometric medianrecursive robust estimationstochastic gradient

Mathematics Subject Classification ID

Nonparametric estimation (62G05) Nonparametric robustness (62G35) Factor analysis and principal components; correspondence analysis (62H25) Stochastic approximation (62L20)

Cites Work

• Title not available (Why is that?)
• Principal component analysis.
• Functional data analysis.
• Robust forecasting of mortality and fertility rates: a functional data approach
• Multivariate location estimation using extension of \(R\)-estimates through \(U\)-statistics type approach
• Robust principal component analysis for functional data. (With comments)
• Linear processes in function spaces. Theory and applications
• Online estimation of the geometric median in Hilbert spaces: nonasymptotic confidence balls
• A comparison of algorithms for the multivariate \(L_1\)-median
• Estimating the geometric median in Hilbert spaces with stochastic gradient algorithms: \(L^p\) and almost sure rates of convergence
• Dispersion operators and resistant second-order functional data analysis
• Robust functional estimation using the median and spherical principal components
• Stochastic Approximation for Multivariate and Functional Median
• Robust Estimation of Dispersion Matrices and Principal Components
• Acceleration of Stochastic Approximation by Averaging
• A Stable and Efficient Algorithm for the Rank-One Modification of the Symmetric Eigenproblem
• Robustifying principal component analysis with spatial sign vectors
• A fast and recursive algorithm for clustering large datasets with \(k\)-medians
• The multivariate L ₁ -median and associated data depth
• Robust functional principal components: a projection-pursuit approach
• Robust Statistics
• High breakdown estimators for principal components: the projection-pursuit approach revis\-ited
• Robust Statistics
• Breakdown points of affine equivariant estimators of multivariate location and covariance matrices
• Convergence rate and averaging of nonlinear two-time-scale stochastic approximation algo\-rithms
• Efficient and fast estimation of the geometric median in Hilbert spaces with an averaged stochastic gradient algorithm
• High-breakdown robust multivariate methods

Cited In (5)

• L^p and almost sure rates of convergence of averaged stochastic gradient algorithms: locally strongly convex objective
• An efficient averaged stochastic Gauss-Newton algorithm for estimating parameters of nonlinear regressions models
• Online estimation of the asymptotic variance for averaged stochastic gradient algorithms
• Robust functional principal components for sparse longitudinal data
• Kmedians

Uses Software

• R
• robustbase
• fda (R)
• Gmedian

This page was built for publication: Fast estimation of the median covariation matrix with application to online robust principal components analysis

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q89458)