Fast rate of convergence in high-dimensional linear discriminant analysis

DOI10.1080/10485252.2010.487531zbMATH Open1359.62240arXiv0909.2191OpenAlexW2082795648MaRDI QIDQ3021183FDOQ3021183

Authors:

Publication date: 22 July 2011

Published in: Journal of Nonparametric Statistics (Search for Journal in Brave)

Abstract: This paper gives a theoretical analysis of high dimensional linear discrimination of Gaussian data. We study the excess risk of linear discriminant rules. We emphasis on the poor performances of standard procedures in the case when dimension p is larger than sample size n. The corresponding theoretical results are non asymptotic lower bounds. On the other hand, we propose two discrimination procedures based on dimensionality reduction and provide associated rates of convergence which can be O(log(p)/n) under sparsity assumptions. Finally all our results rely on a theorem that provides simple sharp relations between the excess risk and an estimation error associated to the geometric parameters defining the used discrimination rule.

Full work available at URL: https://arxiv.org/abs/0909.2191

Recommendations

• The Dantzig discriminant analysis with high dimensional data
• Discriminant analysis of high-dimensional data over limited samples
• High Dimensional Linear Discriminant Analysis: Optimality, Adaptive Algorithm and Missing Data
• L1 least squares for sparse high-dimensional LDA
• Discriminant analysis in small and large dimensions

zbMATH Keywords

classificationdimension reductionlinear discriminant analysishigh dimensionGaussian datathresholding estimator

Mathematics Subject Classification ID

Classification and discrimination; cluster analysis (statistical aspects) (62H30) Asymptotic distribution theory in statistics (62E20) Characterization and structure theory for multivariate probability distributions; copulas (62H05) Estimation in multivariate analysis (62H12)

Cites Work

• Ideal spatial adaptation by wavelet shrinkage
• High-dimensional classification using features annealed independence rules
• Class prediction by nearest shrunken centroids, with applications to DNA microarrays.
• Some theory for Fisher's linear discriminant function, `naive Bayes', and some alternatives when there are many more variables than observations
• Regularized estimation of large covariance matrices
• Adapting to unknown sparsity by controlling the false discovery rate
• Feature selection by higher criticism thresholding achieves the optimal phase diagram
• Innovated higher criticism for detecting sparse signals in correlated noise
• Minimax risk over \(l_ p\)-balls for \(l_ q\)-error
• Adaptive covariance estimation of locally stationary processes
• Smooth discrimination analysis
• Adaptive estimation of a quadratic functional by model selection.
• Minimax estimation via wavelet shrinkage
• Optimal aggregation of classifiers in statistical learning.
• Minimax nonparametric classification .I. Rates of convergence
• Fast learning rates for plug-in classifiers
• Risk bounds for statistical learning
• Modern statistical estimation via oracle inequalities
• Semigroup stationary processes and spectral representation
• Stochastic Deconvolution Over Groups

Cited In (4)

• Dynamic linear discriminant analysis in high dimensional space
• Consistency and convergence rate for nearest subspace classifier
• A theoretical contribution to the fast implementation of null linear discriminant analysis with random matrix multiplication
• On general matrix exponential discriminant analysis methods for high dimensionality reduction

This page was built for publication: Fast rate of convergence in high-dimensional linear discriminant analysis

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3021183)