On the Link Between L1-PCA and ICA

被引:26
作者
Martin-Clemente, Ruben [1 ]
Zarzoso, Vicente [2 ]
机构
[1] Univ Seville, Escuela Super Ingn, Dept Teoria Senal & Comunicac, Avda Descubrimientos S-N, Seville 41092, Spain
[2] Univ Nice Sophia Antipolis, CNRS, Lab I3S, UMR 7271, CS 40121, F-06903 Sophia Antipolis, France
关键词
Feature extraction or construction; interactive data exploration and discovery; independent component analysis; principal component analysis; L1-norm; multivariate statistics; feature representation; feature evaluation and selection; INDEPENDENT COMPONENT ANALYSIS; L1-NORM; ALGORITHMS;
D O I
10.1109/TPAMI.2016.2557797
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Principal component analysis (PCA) based on L1-norm maximization is an emerging technique that has drawn growing interest in the signal processing and machine learning research communities, especially due to its robustness to outliers. The present work proves that L1-norm PCA can perform independent component analysis (ICA) under the whitening assumption. However, when the source probability distributions fulfil certain conditions, the L1-norm criterion needs to be minimized rather than maximized, which can be accomplished by simple modifications on existing optimal algorithms for L1-PCA. If the sources have symmetric distributions, we show in addition that L1-PCA is linked to kurtosis optimization. A number of numerical experiments illustrate the theoretical results and analyze the comparative performance of different algorithms for ICA via L1-PCA. Although our analysis is asymptotic in the sample size, this equivalence opens interesting new perspectives for performing ICA using optimal algorithms for L1-PCA with guaranteed global convergence while inheriting the increased robustness to outliers of the L1-norm criterion.
引用
收藏
页码:515 / 528
页数:14
相关论文
共 32 条
[1]   A polynomial case of unconstrained zero-one quadratic optimization [J].
Allemand, K ;
Fukuda, K ;
Liebling, TM ;
Steiner, E .
MATHEMATICAL PROGRAMMING, 2001, 91 (01) :49-52
[2]  
[Anonymous], 1957, MATH METHODS STAT
[3]   A polynomial-time recursive algorithm for some unconstrained quadratic optimization problems [J].
Ben-Ameur, Walid ;
Neto, Jose .
DISCRETE APPLIED MATHEMATICS, 2011, 159 (16) :1689-1698
[4]   A pure L1-norm principal component analysis [J].
Brooks, J. P. ;
Dula, J. H. ;
Boone, E. L. .
COMPUTATIONAL STATISTICS & DATA ANALYSIS, 2013, 61 :83-98
[5]  
Chong E. K., 2013, INTRO OPTIMIZATION
[6]  
Comon P, 2010, HANDBOOK OF BLIND SOURCE SEPARATION: INDEPENDENT COMPONENT ANALYSIS AND APPLICATIONS, P1
[7]   INDEPENDENT COMPONENT ANALYSIS, A NEW CONCEPT [J].
COMON, P .
SIGNAL PROCESSING, 1994, 36 (03) :287-314
[8]   ADAPTIVE BLIND SEPARATION OF INDEPENDENT SOURCES - A DEFLATION APPROACH [J].
DELFOSSE, N ;
LOUBATON, P .
SIGNAL PROCESSING, 1995, 45 (01) :59-83
[9]  
Ding C, 2006, P 23 INT C MACH LEAR, P281, DOI DOI 10.1145/1143844.1143880
[10]   Solving the fixed rank convex quadratic maximization in binary variables by a parallel zonotope construction algorithm [J].
Ferrez, JA ;
Fukuda, K ;
Liebling, TM .
EUROPEAN JOURNAL OF OPERATIONAL RESEARCH, 2005, 166 (01) :35-50