Independent component analysis for multivariate functional data

Abstract We extend two methods of independent component analysis, fourth order blind identification and joint approximate diagonalization of eigen-matrices, to vector-valued functional data. Multivariate functional data occur naturally and frequently in modern applications, and extending independent component analysis to this setting allows us to distill important information from this type of data, going a step further than the functional principal component analysis. To allow the inversion of the covariance operator we make the assumption that the dependency between the component functions lies in a finite-dimensional subspace. In this subspace we define fourth cross-cumulant operators and use them to construct the two novel, Fisher consistent methods for solving the independent component problem for vector-valued functions. Both simulations and an application on a hand gesture data set show the usefulness and advantages of the proposed methods over functional principal component analysis.

[1]  Erkki Oja,et al.  Efficient Variant of Algorithm FastICA for Independent Component Analysis Attaining the CramÉr-Rao Lower Bound , 2006, IEEE Transactions on Neural Networks.

[2]  Vince D. Calhoun,et al.  ICA of functional MRI data: an overview. , 2003 .

[3]  D. Bosq.  Linear Processes in Function Spaces: Theory And Applications , 2000 .

[4]  Klaus Nordhausen,et al.  Deflation-Based FastICA With Adaptive Choices of Nonlinearities , 2014, IEEE Transactions on Signal Processing.

[5]  Hadley Wickham,et al.  ggplot2 - Elegant Graphics for Data Analysis (2nd Edition) , 2017 .

[6]  Jean-François Cardoso,et al.  Multidimensional independent component analysis , 1998, Proceedings of the 1998 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '98 (Cat. No.98CH36181).

[7]  R Core Team,et al.  R: A language and environment for statistical computing. , 2014 .

[8]  Germain Van Bever,et al.  Functional independent component analysis : an extension of fourth-order blind identification , 2016 .

[9]  Francesca Ieva,et al.  Ecg signal reconstruction, landmark registration and functional classification , 2011 .

[10]  Yoshiharu Sato.  Theoretical Considerations for Multivariate Functional Data Analysis , 2013 .

[11]  Sadanori Konishi,et al.  Functional Cluster Analysis via Orthonormalized Gaussian Basis Expansions and Its Application , 2010, J. Classif..

[12]  Joni Virta,et al.  The squared symmetric FastICA estimator , 2015, Signal Process..

[13]  Hyonho Chun,et al.  On an Additive Semigraphoid Model for Statistical Networks With Application to Pathway Analysis , 2014, Journal of the American Statistical Association.

[14]  Hadley Wickham,et al.  Reshaping Data with the reshape Package , 2007 .

[15]  Marcela Svarc,et al.  Principal components for multivariate functional data , 2011 .

[16]  Joni Virta,et al.  JADE for Tensor-Valued Observations , 2016, Journal of Computational and Graphical Statistics.

[17]  Jean-Francois Cardoso,et al.  Source separation using higher order moments , 1989, International Conference on Acoustics, Speech, and Signal Processing,.

[18]  David S. Matteson,et al.  Likelihood Component Analysis , 2015 .

[19]  C. Baker.  Joint measures and cross-covariance operators , 1973 .

[20]  P. Tichavský,et al.  Efficient variant of algorithm fastica for independent component analysis attaining the cramer-RAO lower bound , 2005, IEEE/SP 13th Workshop on Statistical Signal Processing, 2005.

[21]  D. Chakrabarti,et al.  A fast fixed - point algorithm for independent component analysis , 1997 .

[22]  Julien Jacques,et al.  Model-based clustering for multivariate functional data , 2013, Comput. Stat. Data Anal..

[23]  Klaus Nordhausen,et al.  A New Performance Index for ICA: Properties, Computation and Asymptotic Analysis , 2010, LVA/ICA.

[24]  K. Nordhausen,et al.  Independent Subspace Analysis Using Three Scatter Matrices , 2011 .

[25]  S. Greven,et al.  Multivariate Functional Principal Component Analysis for Data Observed on Different (Dimensional) Domains , 2015, 1509.02029.

[26]  Antoine Souloumiac,et al.  Jacobi Angles for Simultaneous Diagonalization , 1996, SIAM J. Matrix Anal. Appl..

[27]  J. Cardoso,et al.  Blind beamforming for non-gaussian signals , 1993 .

[28]  Hiroshi Yadohisa,et al.  Crisp and fuzzy k-means clustering algorithms for multivariate functional data , 2007, Comput. Stat..

[29]  S. Bonhomme,et al.  Consistent noisy independent component analysis , 2008 .

[30]  B. Li,et al.  Nonlinear sufficient dimension reduction for functional data , 2017 .

[31]  Yoshikazu Terada,et al.  Functional factorial K-means analysis , 2013, Comput. Stat. Data Anal..

[32]  Mia Hubert,et al.  A Measure of Directional Outlyingness With Applications to Image Data and Video , 2016, 1608.05012.

[33]  Eric Moreau,et al.  A generalization of joint-diagonalization criteria for source separation , 2001, IEEE Trans. Signal Process..

[34]  Marc G. Genton,et al.  Multivariate Functional Data Visualization and Outlier Detection , 2017, Journal of Computational and Graphical Statistics.

[35]  Harold W. Gutch,et al.  To Infinity and Beyond: On ICA over Hilbert Spaces , 2012, LVA/ICA.

[36]  Michael I. Jordan,et al.  Kernel independent component analysis , 2003 .

[37]  David S. Matteson,et al.  Linear Non-Gaussian Component Analysis Via Maximum Likelihood , 2015, Journal of the American Statistical Association.

[38]  William N. Venables,et al.  Modern Applied Statistics with S , 2010 .

[39]  B. Li,et al.  A Nonparametric Graphical Model for Functional Data With Application to Brain Networks Based on fMRI , 2018, Journal of the American Statistical Association.

[40]  Jeng-Min Chiou,et al.  Multivariate functional principal component analysis: A normalization approach , 2014 .

[41]  Klaus Nordhausen,et al.  Blind Source Separation Based on Joint Diagonalization in R: The Packages JADE and BSSasymp , 2017 .

[42]  K. Nordhausen,et al.  Fourth Moments and Independent Component Analysis , 2014, 1406.4765.

[43]  Michio Yamamoto,et al.  Clustering of functional data in a low-dimensional subspace , 2012, Advances in Data Analysis and Classification.

[44]  James O. Ramsay,et al.  Functional Data Analysis , 2005 .

[45]  Joni Virta,et al.  Independent component analysis for tensor-valued data , 2016, J. Multivar. Anal..

[46]  Pierre Comon,et al.  Handbook of Blind Source Separation: Independent Component Analysis and Applications , 2010 .

[47]  R. Serfling,et al.  On Invariant Coordinate System (ICS) Functionals , 2012 .

[48]  K. Nordhausen,et al.  New independent component analysis tools for time series , 2015 .

[49]  Zhen Wang,et al.  uWave: Accelerometer-based Personalized Gesture Recognition and Its Applications , 2009, PerCom.

[50]  Jani Mäntyjärvi,et al.  Accelerometer-based gesture control for a design environment , 2006, Personal and Ubiquitous Computing.

[51]  Heungsun Hwang,et al.  Dimension-Reduced Clustering of Functional Data via Subspace Separation , 2017, Journal of Classification.