Robust multivariate L1 principal component analysis and dimensionality reduction

Authors:
Junbin Gao;Paul W. Kwan;Yi Guo
Affiliations:
School of Computer Science, Charles Sturt University, Bathurst, NSW 2795, Australia;School of Science and Technology, University of New England, Armidale, NSW 2351, Australia;School of Science and Technology, University of New England, Armidale, NSW 2351, Australia
Venue:
Neurocomputing
Year:
2009

Citing 10
Cited 2

Robust mixture modelling using the t distribution

Statistics and Computing
A Framework for Robust Subspace Learning

International Journal of Computer Vision - Special Issue on Computational Vision at Brown University
On the Noise Model of Support Vector Machine Regression

On the Noise Model of Support Vector Machine Regression
Feature selection, L1 vs. L2 regularization, and rotational invariance

ICML '04 Proceedings of the twenty-first international conference on Machine learning
Robust L" Norm Factorization in the Presence of Outliers and Missing Data by Alternative Convex Programming

CVPR '05 Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Volume 1 - Volume 01
Robust probabilistic projections

ICML '06 Proceedings of the 23rd international conference on Machine learning
R1-PCA: rotational invariant L1-norm principal component analysis for robust subspace factorization

ICML '06 Proceedings of the 23rd international conference on Machine learning
Robust l1 principal component analysis and its bayesian variational inference

Neural Computation
Twin Kernel Embedding

IEEE Transactions on Pattern Analysis and Machine Intelligence
Variational inference for Student-t models: Robust Bayesian interpolation and generalised component analysis

Neurocomputing

Normalized dimensionality reduction using nonnegative matrix factorization

Neurocomputing
Hybrid structure for robust dimensionality reduction

Neurocomputing

Quantified Score

Hi-index	0.01

Visualization

Abstract

Further to our recent work on the robust L1 PCA we introduce a new version of robust PCA model based on the so-called multivariate Laplace distribution (called L1 distribution) proposed in Eltoft et al. [2006. On the multivariate Laplace distribution. IEEE Signal Process. Lett. 13(5), 300-303]. Due to the heavy tail and high component dependency characteristics of the multivariate L1 distribution, the proposed model is expected to be more robust against data outliers and fitting component dependency. Additionally, we demonstrate how a variational approximation scheme enables effective inference of key parameters in the probabilistic multivariate L1-PCA model. By doing so, a tractable Bayesian inference can be achieved based on the variational EM-type algorithm.