Power spectrum and time-frequency representation (TFR) are two significant methods to analyze EEG data in the frequency and in the time-frequency domain. The nonnegative data in the two domains can be modeled using bilinear and multi-linear transform. In this chapter, the nonnegative matrix factorization (NMF) and tensor decomposition of the canonical polyadic and Tucker models are introduced for decomposing the two-way data and the multi-way data including the modes of time, frequency, space, and subject. The strength of each decomposition model is shown by real EEG data example.
This is a preview of subscription content, log in to check access.
This work was supported by the Fundamental Research Funds for the Central Universities [DUT16JJ(G)03] in Dalian University of Technology, and National Natural Science Foundation of China (Grant No. 81471742).
Becker H, et al. EEG extended source localization: tensor-based vs. conventional methods. Neuroimage. 2014;96(8):143–57.CrossRefPubMedGoogle Scholar
Beckmann CF, Smith SM, Beckmann CF, Smith SM. Tensorial extensions of independent component analysis for multisubject FMRI analysis. Neuroimage. 2005;25(1):294–311.Google Scholar
Bro R. Multi-way analysis in the food industry: theory, algorithms and applications. Doctoral Dissertation: University of Amsterdam; 1998.Google Scholar
Carroll JD, Chang JJ. Analysis of individual differences in multidimensional scaling via an N-way generalization of. Psychometrika. 1970;35(3):283–319.CrossRefGoogle Scholar
Cichocki A. Tensor decompositions: a new concept in brain data analysis? 2013;7(2011).Google Scholar
Cichocki et al. Nonnegative matrix and tensor factorizations. 2009.Google Scholar
Cong F, et al. Extract mismatch negativity and P3a through two-dimensional nonnegative decomposition on time-frequency represented event-related potentials. Berlin/Heidelberg: Springer; 2010. p. 385–91.Google Scholar
Cong F, et al. Benefits of multi-domain feature of mismatch negativity extracted by non-negative tensor factorization from EEG collected by low-density array. Int J Neural Syst. 2012;22(6):1415–28.CrossRefGoogle Scholar
Cong F, et al. Multi-domain feature extraction for small event-related potentials through nonnegative multi-way array decomposition from low dense array EEG. Int J Neural Syst. 2013;23(2):1350006. 10.1142/S0129065713500068Google Scholar
Dempster AP. Maximum likelihood estimation from incomplete data via the EM algorithm (with discussion. in J R Stat Soc Ser B. 1977.Google Scholar
Harshman RA. Foundations of the PARAFAC procedure: model and conditions for an “explanatory” multi-mode factor analysis. In UCLA working papers in. 1969.Google Scholar
Kroonenberg PM. Applied multiway data analysis. Ppsw.rug.nl. 2008.Google Scholar
Lee DD, Seung HS. Learning the parts of objects by non-negative matrix factorization. Nature. 1999;401(6755):788–91.CrossRefPubMedGoogle Scholar
Lee DD, Seung HS. Algorithms for non-negative matrix factorization. In: In NIPS. 2001.Google Scholar
Paatero P. Least squares formulation of robust non-negative factor analysis. Chemom Intell Lab Syst. 1997;37(96):23–35.CrossRefGoogle Scholar
Paatero P, Tapper U. ‡, Positive matrix factorization: a non-negative factor model with optimal utilization of error estimates of data values †. Environmetrics. 1994;5(2):111–26.CrossRefGoogle Scholar
Phan AH, Cichocki A. Extended HALS algorithm for nonnegative Tucker decomposition and its applications for multiway analysis and classification. Neurocomputing. 2011;74(11):1956–69.CrossRefGoogle Scholar
Saul L, Pereira F. Aggregate and mixed-order Markov models for statistical language processing. eprint arXiv:cmp-lg/9706007, 1997;81–9.Google Scholar