Pith. sign in

REVIEW 2 major objections 4 minor 54 references

Beyond Tensor Probabilistic Independent Component Analysis -- Putting Block-Term Decomposition and Independent Vector Analysis Together

T0 review · 2 major / 4 minor · reviewed 2026-07-11 · grok-4.5

Pith's one-line read A BTD-based TPIVA hybrid would unmix realistic multi-subject fMRI sources more reliably than TPICA by replacing collinearity with statistical dependence.

desk verdict A clean, well-cited motivation note that correctly flags the TPICA–BTD–IVA gap but stops short of any model, uniqueness argument, or experiment. read the letter →

arxiv 2607.04272 v1 pith:JD47AS7O submitted 2026-07-05 stat.ME eess.SP

classification stat.MEeess.SP
keywords Block-TermDecomposition(BTD)CanonicalPolyadic(CPD)IndependentVectorAnalysis(IVA)TensorPICA(TPICA)fMRIsourceunmixingsubspaceLL1model
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This note argues that tensor probabilistic ICA (TPICA) is limited for realistic fMRI unmixing because its ICA independence assumption fails when sources overlap strongly in space and its CPD model is too rigid for complex spatial maps. Independent vector analysis better handles dependent components inside independent vectors or subspaces, while block-term decomposition (BTD) already outperforms CPD for the same data. Because algebraic links already exist between IVA and BTD, the author proposes generalizing TPICA into a BTD-based “TPIVA” method, and further generalizing BTD itself so that collinearity is replaced by statistical dependence. A sympathetic reader cares because the hybrid would keep the multi-way structure of multi-subject fMRI while removing the two main failure modes of the currently popular tool, potentially yielding more reproducible source maps under low SNR and variable hemodynamic responses.

What carries the argument

The proposed TPIVA model: a BTD (LL1 or LL11) of the multi-subject fMRI tensor in which each block’s temporal (or subject) factors form an independent vector of mutually dependent components rather than identical collinear vectors, thereby relaxing both the rank-1 CPD constraint and the strict independence of classical ICA.

What would settle it

Construct synthetic multi-subject fMRI tensors with known spatially overlapped rank-(L,L,1) sources whose time courses form independent vectors of controlled dependence; run a concrete TPIVA algorithm and check whether the recovered spatial maps and source-component vectors match the ground truth better than TPICA and pure BTD under the same rank over-estimation and low-SNR conditions.

Watch

Extended reading notes

Core claim

Generalizing the popular TPICA pipeline to a block-term decomposition backbone that incorporates independent vector analysis—termed TPIVA—would more successfully fuse statistical independence assumptions with multi-way tensor structure for multi-subject fMRI source unmixing, especially when spatial maps are overlapped or of rank greater than one; this may also require redefining BTD so that collinear factors become statistically dependent source component vectors.

Load-bearing premise

The known algebraic links between independent vector analysis and block-term decomposition automatically yield a practical, uniquely identifiable hybrid model in which statistical dependence simply replaces collinearity, without new uniqueness or algorithmic obstacles.

Editorial extensions

If this is right

  • Realistic fMRI sources with rich spatial content and subject-specific hemodynamic responses could be unmixed without forcing rank-1 structure or full statistical independence.
  • Model-order selection tools already developed for deterministic BTD could be reused or adapted to estimate both the number of independent vectors and their internal dimensions in a completely blind IVA setting.
  • The same hybrid construction would apply directly to the four-way (space × space × time × subject) tensor obtained by folding the spatial mode, and to online or large-scale variants for dynamic neuroimaging.
  • Analogous source-unmixing problems outside neuroimaging—hyperspectral imaging, radar, communications—could adopt the same dependence-relaxed BTD model.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the soft-dependence generalization of BTD proves identifiable, existing joint block-diagonalization algorithms for cumulant tensors could be repurposed as algebraic initializers for TPIVA, reducing reliance on alternating least-squares iterations.
  • The proposal implicitly suggests a continuum of models between pure CPD and pure IVA; intermediate “shared-subspace” formulations already appearing in the IVA literature could be re-interpreted as low-rank (but not rank-1) BTD factors and tested for uniqueness.
  • Success of TPIVA would also supply a practical route to completely blind independent subspace analysis when both the number of subspaces and their dimensions are unknown a priori.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. The manuscript is a short position note that reviews the known limitations of tensor probabilistic ICA (TPICA) for multi-subject fMRI unmixing—especially spatial source overlap that violates ICA independence and the restrictive rank-1 structure of CPD—and argues that a hybrid “TPIVA” built on block-term decomposition (BTD) and independent vector analysis (IVA) would combine their respective strengths more successfully. After restating the TPICA model (Eqs. 1–2) and the LL1/LL11 BTD models (Eqs. 3–6), Section IV proposes replacing the strict collinearity of the temporal factors inside each block by soft statistical dependence (SCVs of the form (7)–(8)), thereby generalizing both BTD and TPICA. The note surveys related algebraic links between IVA/ISA and BTD, lists possible algorithmic and model-order directions, and sketches future online and multi-modal extensions, but supplies neither a uniqueness argument, an algorithm, nor any numerical illustration for the proposed hybrid.

Significance. If a well-posed, identifiable TPIVA model can be constructed and shown to outperform both TPICA and pure BTD under realistic spatial-overlap and low-SNR conditions, the contribution would be of genuine interest to the multi-subject fMRI and multi-set BSS communities. The manuscript correctly identifies the complementary weaknesses of the two existing paradigms and usefully collates the scattered literature that already points toward their intersection. Because the present text is only a motivating sketch, however, that significance remains prospective rather than demonstrated.

major comments (2)
  1. Section IV (paragraphs surrounding Eqs. (7)–(8)): the central claim that a BTD-based TPIVA “would more successfully combine” statistics and tensors rests on replacing the non-collinearity hypothesis required for LL1 uniqueness (“no null or collinear columns b_r”, §III citing [21]) by soft statistical dependence inside SCVs. The algebraic links cited from Lahat, De Lathauwer et al. concern block-diagonal second-order statistics or irreducible subspaces; they do not automatically guarantee uniqueness or computability once the deterministic rank-(L_r,L_r,1) structure is deliberately relaxed by additive noise Z_r or a low-rank factor U_r V_r^T. Without a fresh uniqueness argument (or at least a clear statement of the open gap), the success claim remains conjectural.
  2. The manuscript contains no algorithm, no model-order selection procedure, and no numerical experiment (synthetic or real) that would allow a reader to assess whether the proposed hybrid is even feasible under the spatial-overlap and low-SNR regimes that already undermine TPICA. For a research-direction note this is not fatal, but the absence of even a minimal proof-of-concept leaves the load-bearing claim untested.
minor comments (4)
  1. Notation for the Khatri-Rao product and the Kruskal operator is introduced without a brief reminder of their definitions; a short glossary or footnote would help non-tensor specialists.
  2. Several acronyms (PPCA, HRF, SCV, ISA, BCA, BCM, DC-CPD, …) appear with only a single expansion; a compact list of abbreviations would improve readability.
  3. The remark that “no more than two iterations are needed” for TPICA is stated without a supporting citation or quantitative reference; a pointer to the original Beckmann–Smith experiments would be useful.
  4. Typographical inconsistencies appear in the rendering of IVA/IV A and in the spacing of multi-letter products (e.g., “IV A”, “TPIV A”).

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: motivational literature survey and research sketch with no derived claims, fits, or self-referential uniqueness reductions.

full rationale

The manuscript is an explicit preliminary note that surveys critiques of TPICA, advantages of BTD over CPD, and known algebraic links between IVA/ISA and BTD (citing external sources such as De Lathauwer, Lahat, Adalı et al.), then sketches a possible hybrid “TPIVA” research direction. It advances no quantitative prediction, no fitted-parameter-to-related-quantity claim, no uniqueness theorem that forces its model choice, and no self-definitional identity. The few self-citations ([9], [25], [26], [51]) report prior empirical observations or algorithmic tools for ordinary BTD; they are not load-bearing premises that close a derivation loop. Softening collinearity to statistical dependence (Eqs. 7–8) is presented as an open modeling idea, not as a result already guaranteed by the cited links. Consequently the derivation chain contains no step that reduces, by the paper’s own equations or by self-citation, to its inputs by construction. Score 0 is the honest finding.

Assumptions & free parameters 0 free parameters · 3 assumptions · 1 invented entities

Because the paper advances no concrete method or theorem, the ledger is almost empty of free parameters and invented entities. The few axioms are standard domain modeling choices already present in the TPICA/BTD/IVA literature it reviews.

assumptions (3)
  • domain assumption fMRI multi-subject data obey a linear mixing model that can be written as a 3-way or 4-way tensor (CPD or BTD).
    Invoked throughout §§I–III as the starting point for both TPICA and the proposed TPIVA.
  • domain assumption Statistical independence (or independence of subspaces/vectors) is a useful and approximately valid assumption for separating fMRI sources.
    Core premise of ICA/IVA that the note both critiques and retains for the hybrid.
  • standard math BTD uniqueness holds under full-column-rank factors and non-collinear (or irreducible) blocks.
    Cited from De Lathauwer’s BTD papers and used to argue that BTD is already identifiable.
invented entities (1)
  • TPIVA (tensor probabilistic independent vector analysis)
    purpose: Name for the still-undefined hybrid of BTD and IVA that would replace TPICA.
    Introduced in the abstract and §IV as the object of future research; no concrete definition, algorithm or uniqueness result is supplied.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Beyond Tensor Probabilistic Independent Component Analysis -- Putting Block-Term Decomposition and Independent Vector Analysis Together." pith.science (2026). https://pith.science/paper/JD47AS7O

@misc{pith2026260704272,
  author       = {Pith},
  title        = {Pith review of: Beyond Tensor Probabilistic Independent Component Analysis -- Putting Block-Term Decomposition and Independent Vector Analysis Together},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JD47AS7O}},
  note         = {Machine review of arXiv:2607.04272}
}
abstract

Tensor probabilistic independent component analysis (TPICA) is a popular approach to analyzing functional magnetic resonance imaging (fMRI) data, which draws its popularity from its ability to enrich the advantages of the statistics-based ICA with the awareness of the multi-way nature of these data, brought about and exploited via a deterministic 3-way (time $\times$ space $\times$ subjects) tensor decomposition (Canonical Polyadic Decomposition (CPD)) model. It has, however, received critique concerning its robustness in realistic fMRI unmixing scenarios, notably those involving sources that are strongly overlapped in space. Such cases may not meet the assumption of statistical independence required in ICA. They can instead be better described as independent vectors (or subspaces) of dependent components, pointing to the adoption of alternative statistical approaches, notably independent vector analysis (IVA). On the other hand, on the deterministic side, CPD is often restrictive and is outperformed by the more flexible block-term decomposition (BTD) model, also in the fMRI source unmixing context. Given the above, plus strong evidence of links between IVA and BTD, it is deemed worthwhile to consider the possibilities of generalizing TPICA to a BTD-based ``TPIVA" extension, which would more successfully combine the power of statistics and tensor decomposition. This could also entail a generalization of the BTD model, where (non)collinearity would be replaced by statistical (in)dependence. This note aims to outline the state-of-the-art and the above ideas in more detail, serving as a preliminary, motivating step in this research direction.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

54 extracted references

  1. [21]

    Decompositions of a higher-order tensor in block terms — Part II: Definitions and uniqueness,

    ——, “Decompositions of a higher-order tensor in block terms — Part II: Definitions and uniqueness,”SIAM J. Matrix Anal. Appl., vol. 30, no. 3, pp. 1033–1066, 2008

  2. [1]

    Tensor decomposition for signal processing and machine learning,

    N. D. Sidiropouloset al., “Tensor decomposition for signal processing and machine learning,”IEEE Trans. Signal Process., vol. 65, no. 13, pp. 3551–3582, Jul. 2017

  3. [2]

    The statistical analysis of fMRI data,

    M. A. Lindquist, “The statistical analysis of fMRI data,”Stat. Sci., vol. 23, no. 4, pp. 439–464, 2008

  4. [3]

    Structure-seeking multilinear methods for the analysis of fMRI data,

    A. H. Andersen and W. S. Rayens, “Structure-seeking multilinear methods for the analysis of fMRI data,”NeuroImage, vol. 22, no. 2, pp. 728–739, 2004

  5. [4]

    Quantifying functional connectivity in multi-subject fMRI data using component models,

    K. H. Madsen, N. W. Churchill, and M. Mørup, “Quantifying functional connectivity in multi-subject fMRI data using component models,”Hum. Brain Mapp., vol. 38, no. 2, pp. 882—-899, 2017

  6. [5]

    Probabilistic independent component analysis for functional magnetic resonance imaging,

    C. F. Beckmann and S. M. Smith, “Probabilistic independent component analysis for functional magnetic resonance imaging,”IEEE Trans. Med. Imag., vol. 23, no. 2, pp. 137–152, Feb. 2004

  7. [6]

    Tensorial extensions of independent component analysis for multisubject fMRI analysis,

    ——, “Tensorial extensions of independent component analysis for multisubject fMRI analysis,”NeuroImage, vol. 25, pp. 294–311, 2005

  8. [7]

    Comparing Independent Component Analysis and the Parafac model for artificial multi-subject fMRI data,

    A. Stegeman, “Comparing Independent Component Analysis and the Parafac model for artificial multi-subject fMRI data,” University of Groningen, Tech. Rep., Feb. 2007

Show all 54 references
  1. [8]

    A critique of Tensor Probabilistic Inde- pendent Component Analysis: Implications and recommendations for multi-subject fMRI data analysis,

    N. E. Helwig and S. Hong, “A critique of Tensor Probabilistic Inde- pendent Component Analysis: Implications and recommendations for multi-subject fMRI data analysis,”J. Neurosci. Meth., vol. 213, no. 2, pp. 263–273, Mar. 2013

  2. [9]

    Blind fMRI source unmixing via higher-order tensor decompositions,

    C. Chatzichristoset al., “Blind fMRI source unmixing via higher-order tensor decompositions,”J. Neurosci. Meth., vol. 315, pp. 17–47, Mar. 2019

  3. [10]

    Independent vector analysis: Definition and algorithms,

    T. Kim, I. Lee, and T.-W. Lee, “Independent vector analysis: Definition and algorithms,” inProc. ACSSC-2006, Pacific Grove, CA, 29 Oct.– 1 Nov. 2006

  4. [11]

    Independent vector analysis (IV A): Multivariate approach for fMRI group study,

    J.-H. Lee, T.-W. Lee, F. A. Jolesz, and S.-S. Yoo, “Independent vector analysis (IV A): Multivariate approach for fMRI group study,”NeuroIm- age, vol. 40, pp. 86–109, 2008

  5. [12]

    Towards a general independent subspace analysis,

    F. J. Theis, “Towards a general independent subspace analysis,” in Proc. NIPS-2006, Vancouver, BC, Canada, Dec. 2006

  6. [13]

    Beyond independent component analysis: Sub- space, coupled, and block decompositions,

    D. Lahat, “Beyond independent component analysis: Sub- space, coupled, and block decompositions,” The Brain Space Initiative Talk Series, 2021. [Online]. Available: https://www.youtube.com/watch?v=WOh2zvj4aWU&t=2s

  7. [14]

    Tensor-based blind fMRI source separation without the Gaussian noise assumption — Aβ-divergence approach,

    C. Chatzichristoset al., “Tensor-based blind fMRI source separation without the Gaussian noise assumption — Aβ-divergence approach,” inProc. GlobalSIP-2019, Ottawa, Canada, Nov. 2019

  8. [15]

    Reproducibility in matrix and tensor decompositions: Focus on model match, interpretability, and uniqueness,

    T. Adalıet al., “Reproducibility in matrix and tensor decompositions: Focus on model match, interpretability, and uniqueness,”IEEE Signal Process. Mag., vol. 39, no. 4, pp. 8–24, Jul. 2022

  9. [16]

    Bayesian inference in fMRI,

    M. W. Woolrich, “Bayesian inference in fMRI,”NeuroImage, vol. 62, pp. 801–810, 2012

  10. [17]

    Tensor regression with applications in neuroimaging data analysis,

    H. Zhou, L. Li, and H. Zhu, “Tensor regression with applications in neuroimaging data analysis,”J. Am. Stat. Assoc., vol. 108, no. 502, pp. 540—-552, 2013

  11. [18]

    CANDECOMP/PARAFAC: From diverging components to a decomposition in block terms,

    A. Stegeman, “CANDECOMP/PARAFAC: From diverging components to a decomposition in block terms,”SIAM J. Matrix Anal. Appl., vol. 33, no. 2, pp. 291—-316, 2012

  12. [19]

    Coherence constrained alternating least squares,

    R. C. Farias, J. H. de Morais Goulart, and P. Comon, “Coherence constrained alternating least squares,” inProc. EUSIPCO-2018, Rome, Italy, Sep. 2018

  13. [20]

    Decompositions of a higher-order tensor in block terms — Part I: Lemmas for partitioned matrices,

    L. De Lathauwer, “Decompositions of a higher-order tensor in block terms — Part I: Lemmas for partitioned matrices,”SIAM J. Matrix Anal. Appl., vol. 30, no. 3, pp. 1022–1032, 2008

  14. [22]

    Decompositions of a higher-order tensor in block terms — Part III: Alternating least squares algorithms,

    L. De Lathauwer and D. Nion, “Decompositions of a higher-order tensor in block terms — Part III: Alternating least squares algorithms,”SIAM J. Matrix Anal. Appl., vol. 30, no. 3, pp. 1067–1083, 2008

  15. [23]

    Block component analysis: a new concept for blind source separation,

    L. De Lathauwer, “Block component analysis: a new concept for blind source separation,” inProc. LVA/ICA-2012, Tel Aviv, Israel, Mar. 2012

  16. [24]

    A tensor-based method for large-scale blind source separation using segmentation,

    M. Bouss ´e, O. Debals, and L. De Lathauwer, “A tensor-based method for large-scale blind source separation using segmentation,”IEEE Trans. Signal Process., vol. 65, no. 2, pp. 346–358, Jan. 2017

  17. [25]

    Block-term tensor decomposition: Model selection and computation,

    A. A. Rontogiannis, E. Kofidis, and P. V . Giampouras, “Block-term tensor decomposition: Model selection and computation,”IEEE J. Sel. Topics Signal Process., vol. 15, no. 3, pp. 464–475, Apr. 2021

  18. [26]

    Block-term tensor decomposition model selection and computation: The Bayesian way,

    P. V . Giampouras, A. A. Rontogiannis, and E. Kofidis, “Block-term tensor decomposition model selection and computation: The Bayesian way,”IEEE Trans. Signal Process., vol. 70, pp. 1704–1717, 2022

  19. [27]

    Tensor and coupled decompositions in block terms: Uniqueness and irreducibility,

    D. Lahat and C. Jutten, “Tensor and coupled decompositions in block terms: Uniqueness and irreducibility,” inProc. SPARS-2019, Toulouse, France, Jul. 2019

  20. [28]

    A combination of parallel factor and independent component analysis,

    M. De V oset al., “A combination of parallel factor and independent component analysis,”Signal Process., vol. 92, no. 12, pp. 2990–2999, Dec. 2012

  21. [29]

    Blind deconvolution of DS- CDMA signals by means of decomposition in rank-(1, L, L)terms,

    L. De Lathauwer and A. De Baynast, “Blind deconvolution of DS- CDMA signals by means of decomposition in rank-(1, L, L)terms,” IEEE Trans. Signal Process., vol. 56, no. 4, pp. 1562–1571, Apr. 2008

  22. [30]

    Tensor clustering on outer-product of coefficient and component matrices of independent component analysis for reliable functional magnetic resonance imaging data decomposition,

    G. Huet al., “Tensor clustering on outer-product of coefficient and component matrices of independent component analysis for reliable functional magnetic resonance imaging data decomposition,”J. Neu- rosci. Meth., vol. 325, no. 108359, Sep. 2019

  23. [31]

    A scalable approach to independent vector analysis by shared subspace separation for multi-subject fMRI analysis,

    M. Sunet al., “A scalable approach to independent vector analysis by shared subspace separation for multi-subject fMRI analysis,”Sensors, vol. 23, 2023

  24. [32]

    An efficient analytic solution for joint blind source separation,

    B. Gabrielsonet al., “An efficient analytic solution for joint blind source separation,”IEEE Trans. Signal Process., vol. 72, pp. 2436–2449, May 2024

  25. [33]

    Early soft and flexible fusion of EEG and fMRI via Double CMTF for multi-subject group analysis,

    C. Chatzichristoset al., “Early soft and flexible fusion of EEG and fMRI via Double CMTF for multi-subject group analysis,”Hum. Brain Mapp., vol. 43, no. 4, pp. 1231—-1255, 2022

  26. [34]

    Spatial and temporal independent component anal- ysis of functional MRI data containing a pair of task-related waveforms,

    V . D. Calhounet al., “Spatial and temporal independent component anal- ysis of functional MRI data containing a pair of task-related waveforms,” Hum. Brain Mapp., vol. 13, no. 1, pp. 43–53, May 2001

  27. [35]

    Performance of temporal and spatial independent component analysis in identifying and removing low- frequency physiological and motion effects in resting-state fMRI,

    A. M. Golestani and J. J. Chen, “Performance of temporal and spatial independent component analysis in identifying and removing low- frequency physiological and motion effects in resting-state fMRI,”Front. Neurosci., vol. 16, Jun. 2022

  28. [36]

    A unified framework for group independent component analysis for multi-subject fMRI data,

    Y . Guo and G. Pagnoni, “A unified framework for group independent component analysis for multi-subject fMRI data,”NeuroImage, vol. 42, no. 3, pp. 1078–1093, 2008

  29. [37]

    Linked independent component analysis for multimodal data fusion,

    A. R. Groveset al., “Linked independent component analysis for multimodal data fusion,”NeuroImage, vol. 54, p. 2198–2217, 2011

  30. [38]

    Independent component analysis for three-way data with an application from atmospheric science,

    S. Unkelet al., “Independent component analysis for three-way data with an application from atmospheric science,”J. Agric. Biol. Env. Stat., vol. 16, no. 3, pp. 319—-338, 2011

  31. [39]

    A blind block term decomposition of high order tensors,

    Y . Cai and P. Li, “A blind block term decomposition of high order tensors,” inProc. AAAI-2021, Feb. 2021

  32. [40]

    On uniqueness and computation of the decomposition of a tensor into multilinear rank-(1, L r, Lr)terms,

    I. Domanov and L. De Lathauwer, “On uniqueness and computation of the decomposition of a tensor into multilinear rank-(1, L r, Lr)terms,” SIAM J. Matrix Anal. Appl., vol. 41, no. 2, 2020

  33. [41]

    Joint independent subspace analysis: Unique- ness and identifiability,

    D. Lahat and C. Jutten, “Joint independent subspace analysis: Unique- ness and identifiability,”IEEE Trans. Signal Process., vol. 67, no. 3, pp. 684–699, Feb. 2019

  34. [42]

    A new link between joint blind source separation using sec- ond order statistics and the canonical polyadic decomposition,

    ——, “A new link between joint blind source separation using sec- ond order statistics and the canonical polyadic decomposition,” in Proc. LVA/ICA-2018, Guilford, UK, Jul. 2018

  35. [43]

    Double coupled canonical polyadic decomposition for joint blind source separation,

    X.-F. Gonget al., “Double coupled canonical polyadic decomposition for joint blind source separation,”IEEE Trans. Signal Process., vol. 66, no. 13, pp. 3475–3490, Jul. 2018

  36. [44]

    Double coupled canonical polyadic decomposition of third-order tensors: Algebraic algorithm and relaxed uniqueness conditions,

    ——, “Double coupled canonical polyadic decomposition of third-order tensors: Algebraic algorithm and relaxed uniqueness conditions,”Signal Process.: Image Commun., vol. 73, pp. 22–36, 2019

  37. [45]

    A tensor framework for nonunitary joint block diagonaliza- tion,

    D. Nion, “A tensor framework for nonunitary joint block diagonaliza- tion,”IEEE Trans. Signal Process., vol. 59, no. 10, pp. 4585–4594, Oct. 2011

  38. [46]

    Multi-task fMRI data fusion using IV A and PARAFAC2,

    I. Lehmannet al., “Multi-task fMRI data fusion using IV A and PARAFAC2,” inProc. ICASSP-2022, Singapore, May 2022

  39. [47]

    An explicit connection between independent vector analysis and tensor decomposition in blind source separation,

    H. Ruanet al., “An explicit connection between independent vector analysis and tensor decomposition in blind source separation,”IEEE Signal Process. Lett., vol. 29, pp. 1277–1281, May 2022

  40. [48]

    Independent vector analysis: Model, applications, challenges,

    Z. Luo, “Independent vector analysis: Model, applications, challenges,” Pattern Recognit., vol. 138, no. 109376, 2023

  41. [49]

    Dynamic independent component/vector analysis: Time-variant linear mixtures separable by time-invariant beamformers,

    Z. Koldovsk ´yet al., “Dynamic independent component/vector analysis: Time-variant linear mixtures separable by time-invariant beamformers,” IEEE Trans. Signal Process., vol. 69, pp. 2158 – 2173, Mar. 2021

  42. [50]

    Time-varying brain connectivity in fMRI data: Whole-brain data-driven approaches for capturing and character- izing dynamic states,

    V . D. Calhoun and T. Adalı, “Time-varying brain connectivity in fMRI data: Whole-brain data-driven approaches for capturing and character- izing dynamic states,”IEEE Signal Process. Mag., vol. 33, no. 3, pp. 52–66, May 2016

  43. [51]

    Online rank- revealing block-term tensor decomposition,

    A. A. Rontogiannis, E. Kofidis, and P. V . Giampouras, “Online rank- revealing block-term tensor decomposition,”Signal Process., vol. 212, no. 109126, Nov. 2023

  44. [52]

    Independent vector analysis with multivariate Gaussian model: A scalable method by multilinear regression,

    B. Gabrielsonet al., “Independent vector analysis with multivariate Gaussian model: A scalable method by multilinear regression,” in Proc. ICASSP-2023, Rhodes, Greece, Jun. 2023

  45. [53]

    Robust tensor-based techniques for antenna array- based GNSS receivers in scenarios with highly correlated multipath components,

    D. V . de Limaet al., “Robust tensor-based techniques for antenna array- based GNSS receivers in scenarios with highly correlated multipath components,”Digit. Signal Process., vol. 101, Jun. 2020

  46. [54]

    Independent vector analysis for blindly deconvolving digital modulated communication signals,

    Z. Luo, R. Guo, and C. Li, “Independent vector analysis for blindly deconvolving digital modulated communication signals,”Electronics, vol. 11, no. 1460, 2022

Pith tools

Reviewed July 11, 2026 · model on record in the stance chip above.