pith. machine review for the scientific record. sign in

arxiv: 2605.13312 · v1 · submitted 2026-05-13 · 💻 cs.LG

Recognition: 3 theorem links

· Lean Theorem

Supervised Deep Multimodal Matrix Factorization for Interpretable Brain Network Analysis

Amjad Seyedi , Lifang He , Songlin Zhao , Akwum Onwunta , Nicolas Gillis
Authors on Pith no claims yet

Pith reviewed 2026-05-14 19:14 UTC · model grok-4.3

classification 💻 cs.LG
keywords brain network analysismultimodal matrix factorizationinterpretable machine learningsupervised graph learningconnectome datacommunity detectiondeep factorization
0
0 comments X

The pith

Supervised deep multimodal matrix factorization learns interpretable community structures from brain graphs for better prediction.

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

The paper introduces SD3MF to extend symmetric nonnegative matrix tri-factorization from single-graph unsupervised clustering to supervised prediction across populations of multimodal brain graphs. It builds hierarchical factorizations separately for each modality while learning a shared latent space that aligns subjects across views. An encoder-decoder setup jointly optimizes reconstruction of the input graphs and the supervised prediction task, with adaptive weights handling fusion of the modalities in a data-driven way. Subjects are represented by community-level interaction matrices that serve as both predictive features and biologically readable summaries. On multimodal connectome datasets the approach outperforms CNN and GNN baselines while surfacing interpretable network patterns.

Core claim

SD3MF learns deep hierarchical factorizations for each modality together with a shared latent representation that aligns subjects across views. An encoder-decoder formulation jointly optimizes graph reconstruction and supervised prediction, while adaptive weights enable data-driven multimodal fusion. By representing each subject through community-level interaction matrices, the model yields interpretable and discriminative features.

What carries the argument

SD3MF, the supervised extension of symmetric nonnegative matrix tri-factorization that produces community-level interaction matrices for each subject while aligning multimodal views through a shared latent space.

If this is right

  • The model outperforms strong deep learning baselines such as CNNs and GNNs on multimodal connectome datasets.
  • Community interaction matrices provide both accurate predictions and biologically interpretable insights.
  • Adaptive weights perform data-driven fusion across modalities without manual tuning.
  • Hierarchical factorizations per modality allow the model to capture structure at multiple scales.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The shared latent representation could support transfer or joint modeling across additional brain imaging modalities not used in the original training.
  • Community matrices might serve as compact biomarkers for tracking disease progression if applied to longitudinal connectome data.
  • The same factorization approach could be tested on other multimodal graph domains such as social or transportation networks where interpretability of communities matters.

Load-bearing premise

The community-level interaction matrices are assumed to capture biologically meaningful and discriminative structure without external validation against known brain atlases or functional networks.

What would settle it

An experiment showing that the learned community matrices have no statistically significant overlap with established brain atlases or known functional networks, or that removing them does not degrade prediction performance, would falsify the interpretability and utility claims.

Figures

Figures reproduced from arXiv: 2605.13312 by Akwum Onwunta, Amjad Seyedi, Lifang He, Nicolas Gillis, Songlin Zhao.

Figure 1
Figure 1. Figure 1: SD3MF architecture: The decoder (top) reconstructs each of the three modality-specific [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Sensitivity vs. speci￾ficity of top methods. 60 70 80 90 Specificity 60 70 80 90 100 Method Sensitivity SGCN TGNet SD3MF Dataset HIV BP PPMI Results [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Influence of the regularization parameter [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Shallow (S3MF) vs. deep (SD3MF) models in terms of ACC and AUC. HIV BP PPMI 60 70 80 90 ACC HIV BP PPMI 60 70 80 90 100 AUC Shallow (S3MF) model Deep (SD3MF) model Interpretation of Learned Representations. We provide an intrinsic interpretability analysis by linking learned parameters to neurobiologically meaningful community structure and discriminative regional patterns. Specifically, we interpret (i) c… view at source ↗
Figure 5
Figure 5. Figure 5: Interpretation of learned communities for the HIV cohort. ROIs are grouped by latent [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: Surface visualization of modality￾specific and multimodal salient ROIs for distin￾guishing HIV from healthy controls. Results are rendered on lateral, medial, and ventral views. DTI-derived communities more prominently co-localize subcortical nuclei with frontal and temporal regions, consistent with the known sensitivity of diffusion MRI to fronto-subcortical and white-matter related alterations in HIV [62… view at source ↗
Figure 8
Figure 8. Figure 8: Convergence behavior of the proposed SD3MF with learning rate [PITH_FULL_IMAGE:figures/full_fig_p016_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Visualization of the learned membership matrices [PITH_FULL_IMAGE:figures/full_fig_p018_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Approximation matrices Ab (m) HIV = Ψ(m)S¯HIVΨ(m)⊤ in ROI space for the HIV cohort under DTI, fMRI, and multimodal settings. Entries represent approximated ROI-to-ROI connectivity strengths. ROIs are reordered by dominant community assignments (C1–C10) shown on the axes for visualization. E Learned Membership Matrix [PITH_FULL_IMAGE:figures/full_fig_p018_10.png] view at source ↗
read the original abstract

We present Supervised Deep Multimodal Matrix Factorization (SD3MF), an interpretable framework for integrative brain network analysis that generalizes Symmetric Nonnegative Matrix Tri-Factorization (SNMTF) from unsupervised single-graph clustering to supervised prediction over populations of multimodal graphs. SD3MF learns deep hierarchical factorizations for each modality together with a shared latent representation that aligns subjects across views. An encoder-decoder formulation jointly optimizes graph reconstruction and supervised prediction, while adaptive weights enable data-driven multimodal fusion. By representing each subject through community-level interaction matrices, the model yields interpretable and discriminative features. Experiments on multimodal connectome datasets show that SD3MF consistently outperforms strong deep learning baselines such as CNNs and GNNs, while enabling biologically interpretable insights. Code for reproducibility is available at: https://github.com/amjadseyedi/SD3MF.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper introduces Supervised Deep Multimodal Matrix Factorization (SD3MF), a generalization of Symmetric Nonnegative Matrix Tri-Factorization to a supervised deep hierarchical setting for multimodal brain connectome analysis. It employs an encoder-decoder architecture that jointly optimizes graph reconstruction and subject-level prediction losses, with adaptive multimodal fusion and community-level interaction matrices as the interpretable subject representations. The central claims are consistent outperformance over CNN and GNN baselines on multimodal connectome datasets together with biologically interpretable insights derived from the learned community structures.

Significance. If the empirical and interpretability claims hold, SD3MF would supply a parameter-efficient, interpretable alternative to black-box deep models for integrative neuroimaging, with the community interaction matrices offering a potential bridge to neuroscientific analysis. The public code release at https://github.com/amjadseyedi/SD3MF supports reproducibility and is a clear strength.

major comments (2)
  1. [Abstract] Abstract: the claim that SD3MF 'consistently outperforms strong deep learning baselines such as CNNs and GNNs' is presented without any quantitative results, statistical tests, data-split details, or baseline specifications; this absence renders the primary empirical contribution impossible to assess and is load-bearing for the paper's central claim.
  2. [Abstract] Abstract: the assertion that community-level interaction matrices 'yield interpretable and discriminative features' and enable 'biologically interpretable insights' rests on the unvalidated assumption that these matrices recover or align with established neuroscientific structures (e.g., AAL, Desikan-Killiany, Yeo 7/17 networks); no alignment metrics, atlas comparisons, or external validation are described, undermining the interpretability half of the contribution.
minor comments (2)
  1. [Abstract] The abstract states that the model 'generalizes Symmetric Nonnegative Matrix Tri-Factorization (SNMTF)' but does not specify which components of SNMTF are retained versus modified in the deep supervised extension; a brief equation-level comparison would clarify the novelty.
  2. [Abstract] The phrase 'adaptive weights enable data-driven multimodal fusion' is used without indicating whether these weights are learned end-to-end or set by a separate procedure; notation for the fusion mechanism should be introduced explicitly.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive feedback, which has helped clarify the presentation of our empirical and interpretability claims. We address each major comment point by point below, indicating the revisions we will make to the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that SD3MF 'consistently outperforms strong deep learning baselines such as CNNs and GNNs' is presented without any quantitative results, statistical tests, data-split details, or baseline specifications; this absence renders the primary empirical contribution impossible to assess and is load-bearing for the paper's central claim.

    Authors: We agree that the abstract would benefit from including key quantitative support to substantiate the performance claims. In the revised manuscript, we have updated the abstract to briefly report mean classification accuracy (with standard deviation) across 5-fold cross-validation, p-values from paired statistical tests against the baselines, the exact data-split protocol, and the specific CNN and GNN architectures used (including layer counts and hyperparameters). Full experimental details and tables remain in Sections 4 and 5. revision: yes

  2. Referee: [Abstract] Abstract: the assertion that community-level interaction matrices 'yield interpretable and discriminative features' and enable 'biologically interpretable insights' rests on the unvalidated assumption that these matrices recover or align with established neuroscientific structures (e.g., AAL, Desikan-Killiany, Yeo 7/17 networks); no alignment metrics, atlas comparisons, or external validation are described, undermining the interpretability half of the contribution.

    Authors: We thank the referee for this observation. The current manuscript supports interpretability through qualitative visualizations of the learned community structures and their contribution to subject-level prediction. To address the lack of quantitative validation, we have added a new analysis subsection that reports alignment metrics (normalized mutual information and Dice coefficients) between the discovered communities and the Yeo 7/17 networks, along with comparisons to AAL and Desikan-Killiany parcellations. These results are now summarized in the abstract and detailed in the revised results section. revision: yes

Circularity Check

0 steps flagged

No derivation circularity; model is defined by independent optimization and experiments

full rationale

The paper presents SD3MF as a new encoder-decoder formulation that jointly optimizes graph reconstruction and supervised prediction losses, generalizing SNMTF. No equation or step reduces a claimed prediction or interpretability result to a fitted parameter by construction. The outperformance claim rests on experimental comparisons to CNNs and GNNs rather than algebraic identity. The biological interpretability of community matrices is asserted without external validation, but this is an unverified assumption, not a circular reduction in the derivation chain. No load-bearing self-citation or ansatz smuggling is exhibited in the provided text. This is the normal case of a self-contained model definition.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The framework rests on nonnegativity constraints for interpretability, the existence of meaningful community structure in brain graphs, and the assumption that joint reconstruction-plus-prediction optimization yields aligned multimodal representations.

free parameters (2)
  • number of communities / latent factors
    Chosen per modality to represent brain communities; value not specified in abstract.
  • depth of hierarchical factorization
    Number of layers in the deep factorization; selected to balance expressivity and interpretability.
axioms (2)
  • domain assumption Nonnegative factors yield parts-based interpretable representations
    Standard assumption in nonnegative matrix factorization invoked to justify community-level interpretability.
  • domain assumption Multimodal brain graphs share a common latent subject space
    Required for the shared representation to align subjects across views.

pith-pipeline@v0.9.0 · 5456 in / 1313 out tokens · 79943 ms · 2026-05-14T19:14:12.780300+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

78 extracted references · 78 canonical work pages · 2 internal anchors

  1. [1]

    Leow, Paul M

    Haoteng Tang, Guixiang Ma, Yanfu Zhang, Kai Ye, Lei Guo, Guodong Liu, Qi Huang, Yalin Wang, Olusola Ajilore, Alex D. Leow, Paul M. Thompson, Heng Huang, and Liang Zhan. A comprehensive survey of complex brain network representation.Meta-Radiology, 1(3):100046, 2023

    work page 2023

  2. [2]

    Haijing Sun, Anna Wang, and Shanshan He. Temporal and spatial analysis of Alzheimer’s disease based on an improved convolutional neural network and a resting-state FMRI brain functional network.International Journal of Environmental Research and Public Health, 19(8): 4508, 2022

    work page 2022

  3. [3]

    Chen, Yu Zhang, and Lifang He

    Rong Zhou, Houliang Zhou, Li Shen, Brian Y . Chen, Yu Zhang, and Lifang He. Integrating multimodal contrastive learning and cross-modal attention for Alzheimer’s disease prediction in brain imaging genetics. InIEEE International Conference on Bioinformatics and Biomedicine (BIBM), pages 1806–1811, 2023

    work page 2023

  4. [4]

    Multimodal neuroimaging: basic concepts and classification of neuropsychiatric diseases.Clinical EEG and Neuroscience, 50(1):20–33, 2019

    Emine Elif Tulay, Barı¸ s Metin, Nevzat Tarhan, and Mehmet Kemal Arıkan. Multimodal neuroimaging: basic concepts and classification of neuropsychiatric diseases.Clinical EEG and Neuroscience, 50(1):20–33, 2019

    work page 2019

  5. [5]

    Multimodal neuroimaging computing: a review of the applications in neuropsychiatric disorders.Brain Informatics, 2(3):167–180, 2015

    Sidong Liu, Weidong Cai, Siqi Liu, Fan Zhang, Michael Fulham, Dagan Feng, Sonia Pujol, and Ron Kikinis. Multimodal neuroimaging computing: a review of the applications in neuropsychiatric disorders.Brain Informatics, 2(3):167–180, 2015

    work page 2015

  6. [6]

    Deep generalized canonical correlation analysis

    Adrian Benton, Huda Khayrallah, Biman Gujral, Dee Ann Reisinger, Sheng Zhang, and Raman Arora. Deep generalized canonical correlation analysis. InWorkshop on Representation Learning for NLP, pages 1–6, 2019

    work page 2019

  7. [7]

    Attentive deep canonical correlation analysis for diagnosing Alzheimer’s disease using multimodal imaging genetics

    Rong Zhou, Houliang Zhou, Brian Y Chen, Li Shen, Yu Zhang, and Lifang He. Attentive deep canonical correlation analysis for diagnosing Alzheimer’s disease using multimodal imaging genetics. InInternational Conference on Medical Image Computing and Computer-Assisted Intervention, pages 681–691. Springer, 2023

    work page 2023

  8. [8]

    Shiju, and S

    Asif Salim, S.S. Shiju, and S. Sumitra. Design of multi-view graph embedding using multiple kernel learning.Engineering Applications of Artificial Intelligence, 90:103534, 2020

    work page 2020

  9. [9]

    Stephen, Tony W

    Yipu Zhang, Li Xiao, Gemeng Zhang, Biao Cai, Julia M. Stephen, Tony W. Wilson, Vince D. Calhoun, and Yu-Ping Wang. Multi-paradigm fMRI fusion via sparse tensor decomposition in brain functional connectivity study.IEEE Journal of Biomedical and Health Informatics, 25(5): 1712–1723, 2021

    work page 2021

  10. [10]

    Wilson, Vince D

    Irina Belyaeva, Ben Gabrielson, Yu-Ping Wang, Tony W. Wilson, Vince D. Calhoun, Julia M. Stephen, and Tülay Adali. Learning spatiotemporal brain dynamics in adolescents via multi- modal MEG and fMRI data fusion using joint tensor/matrix decomposition.IEEE Transactions on Biomedical Engineering, 71(7):2189–2200, 2024. 10

    work page 2024

  11. [11]

    Boosted sparse and low-rank tensor regression

    Lifang He, Kun Chen, Wanwan Xu, Jiayu Zhou, and Fei Wang. Boosted sparse and low-rank tensor regression. InAdvances in Neural Information Processing Systems, volume 31, 2018

    work page 2018

  12. [12]

    Structural deep brain network mining

    Shen Wang, Lifang He, Bokai Cao, Chun-Ta Lu, Philip S Yu, and Ann B Ragin. Structural deep brain network mining. InACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pages 475–484, 2017

    work page 2017

  13. [13]

    Clustering-based deep brain multigraph integrator network for learning connectional brain templates

    U˘gur Demir, Mohammed Amine Gharsallaoui, and Islem Rekik. Clustering-based deep brain multigraph integrator network for learning connectional brain templates. InInternational Workshop on Uncertainty for Safe Utilization of Machine Learning in Medical Imaging, pages 109–120, 2020

    work page 2020

  14. [14]

    Discriminative analysis of schizophrenia patients using graph convolutional networks: A combined multimodal MRI and connectomics analysis

    Xiaoyi Chen, Pengfei Ke, Yuanyuan Huang, Jing Zhou, Hehua Li, Runlin Peng, Jiayuan Huang, Liqin Liang, Guolin Ma, Xiaobo Li, et al. Discriminative analysis of schizophrenia patients using graph convolutional networks: A combined multimodal MRI and connectomics analysis. Frontiers in Neuroscience, 17:1140801, 2023

    work page 2023

  15. [15]

    A generalist vision–language foundation model for diverse biomedical tasks.Nature Medicine, 30(11):3129–3141, 2024

    Kai Zhang, Rong Zhou, Eashan Adhikarla, Zhiling Yan, Yixin Liu, Jun Yu, Zhengliang Liu, Xun Chen, Brian D Davison, Hui Ren, et al. A generalist vision–language foundation model for diverse biomedical tasks.Nature Medicine, 30(11):3129–3141, 2024

    work page 2024

  16. [16]

    Multi-view graph convolutional network and its applications on neuroimage analysis for Parkinson’s disease

    Xi Zhang, Lifang He, Kun Chen, Yuan Luo, Jiayu Zhou, and Fei Wang. Multi-view graph convolutional network and its applications on neuroimage analysis for Parkinson’s disease. In AMIA annual symposium proceedings, volume 2018, page 1147, 2018

    work page 2018

  17. [17]

    MVS- GCN: A prior brain structure learning-guided multi-view graph convolution network for autism spectrum disorder diagnosis.Computers in Biology and Medicine, 142:105239, 2022

    Guangqi Wen, Peng Cao, Huiwen Bao, Wenju Yang, Tong Zheng, and Osmar Zaiane. MVS- GCN: A prior brain structure learning-guided multi-view graph convolution network for autism spectrum disorder diagnosis.Computers in Biology and Medicine, 142:105239, 2022

    work page 2022

  18. [18]

    Calhoun, Aiying Zhang, and Yu-Ping Wang

    Gang Qu, Ziyu Zhou, Vince D. Calhoun, Aiying Zhang, and Yu-Ping Wang. Integrated brain connectivity analysis with fMRI, DTI, and sMRI powered by interpretable graph neural networks.Medical Image Analysis, 103:103570, 2025

    work page 2025

  19. [19]

    Chen, Li Shen, and Yu Zhang

    Houliang Zhou, Lifang He, Brian Y . Chen, Li Shen, and Yu Zhang. Multi-modal diagnosis of Alzheimer’s disease using interpretable graph convolutional networks.IEEE Transactions on Medical Imaging, 44(1):142–153, 2025

    work page 2025

  20. [20]

    Multi- view brain networks construction for Alzheimer’s disease diagnosis

    Yuefeng Ma, Tengfei Zhang, Zeqi Wu, Xiaochen Mu, Xun Liang, and Lanzhen Guo. Multi- view brain networks construction for Alzheimer’s disease diagnosis. InIEEE International Conference on Bioinformatics and Biomedicine (BIBM), pages 889–892, 2023

    work page 2023

  21. [21]

    Plataniotis, and Anastasios N

    Haiping Lu, Konstantinos N. Plataniotis, and Anastasios N. Venetsanopoulos. Mpca: Multilinear principal component analysis of tensor objects.IEEE Transactions on Neural Networks, 19(1): 18–39, 2008

    work page 2008

  22. [22]

    Joint embedding of structural and functional brain networks with graph neural networks for mental illness diagnosis

    Yanqiao Zhu, Hejie Cui, Lifang He, Lichao Sun, and Carl Yang. Joint embedding of structural and functional brain networks with graph neural networks for mental illness diagnosis. In International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), pages 272–276, 2022

    work page 2022

  23. [23]

    Clustering on multi-source incomplete data via tensor modeling and factorization

    Weixiang Shao, Lifang He, and Philip S Yu. Clustering on multi-source incomplete data via tensor modeling and factorization. InPacific-Asia Conference on Knowledge Discovery and Data Mining, pages 485–497. Springer, 2015

    work page 2015

  24. [24]

    Multiview subspace clustering via tensorial t-product representation.IEEE Transactions on Neural Networks and Learning Systems, 30(3): 851–864, 2019

    Ming Yin, Junbin Gao, Shengli Xie, and Yi Guo. Multiview subspace clustering via tensorial t-product representation.IEEE Transactions on Neural Networks and Learning Systems, 30(3): 851–864, 2019

    work page 2019

  25. [25]

    Yu, Alex D

    Guixiang Ma, Lifang He, Chun-Ta Lu, Weixiang Shao, Philip S. Yu, Alex D. Leow, and Ann B. Ragin. Multi-view clustering with graph embedding for connectome analysis. InProceedings of the 2017 ACM on Conference on Information and Knowledge Management, page 127–136, 2017. 11

    work page 2017

  26. [26]

    Yu, Ann B

    Ye Liu, Lifang He, Bokai Cao, Philip S. Yu, Ann B. Ragin, and Alex D. Leow. Multi-view multi- graph embedding for brain network clustering analysis. InProceedings of the Thirty-Second AAAI Conference on Artificial Intelligence, 2018

    work page 2018

  27. [27]

    Fully automated diagnosis of thyroid nodule ultrasound using brain-inspired inference.Neurocomputing, 582:127497, 2024

    Guanghui Li, Qinghua Huang, Chunying Liu, Guanying Wang, Lingli Guo, Ruonan Liu, and Longzhong Liu. Fully automated diagnosis of thyroid nodule ultrasound using brain-inspired inference.Neurocomputing, 582:127497, 2024

    work page 2024

  28. [28]

    Brown, Steven P

    Jeremy Kawahara, Colin J. Brown, Steven P. Miller, Brian G. Booth, Vann Chau, Ruth E. Grunau, Jill G. Zwicker, and Ghassan Hamarneh. Brainnetcnn: Convolutional neural networks for brain networks; towards predicting neurodevelopment.NeuroImage, 146:1038–1049, 2017

    work page 2017

  29. [29]

    Semi-Supervised Classification with Graph Convolutional Networks

    Thomas N. Kipf and Max Welling. Semi-supervised classification with graph convolutional networks, 2017. URLhttps://arxiv.org/abs/1609.02907

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  30. [30]

    Inductive representation learning on large graphs

    Will Hamilton, Zhitao Ying, and Jure Leskovec. Inductive representation learning on large graphs. InAdvances in Neural Information Processing Systems, volume 30, 2017

    work page 2017

  31. [31]

    Graph convolution based attention model for personalized disease prediction

    Anees Kazi, Shayan Shekarforoush, S Arvind Krishna, Hendrik Burwinkel, Gerome Vivar, Benedict Wiestler, Karsten Kortüm, Seyed-Ahmad Ahmadi, Shadi Albarqouni, and Nassir Navab. Graph convolution based attention model for personalized disease prediction. In International Conference on Medical Image Computing and Computer-Assisted Intervention, pages 122–130...

    work page 2019

  32. [32]

    Mapping multi- modal brain connectome for brain disorder diagnosis via cross-modal mutual learning.IEEE Transactions on Medical Imaging, 43(1):108–121, 2024

    Yanwu Yang, Chenfei Ye, Xutao Guo, Tao Wu, Yang Xiang, and Ting Ma. Mapping multi- modal brain connectome for brain disorder diagnosis via cross-modal mutual learning.IEEE Transactions on Medical Imaging, 43(1):108–121, 2024

    work page 2024

  33. [33]

    Rajapakse

    Yanteng Zhang, Xiaohai He, Yi Hao Chan, Qizhi Teng, and Jagath C. Rajapakse. Multi-modal graph neural network for early diagnosis of Alzheimer’s disease from sMRI and PET scans. Computers in Biology and Medicine, 164:107328, 2023

    work page 2023

  34. [34]

    Interpretable graph neural networks for connectome-based brain disorder analysis

    Hejie Cui, Wei Dai, Yanqiao Zhu, Xiaoxiao Li, Lifang He, and Carl Yang. Interpretable graph neural networks for connectome-based brain disorder analysis. InInternational Conference on Medical Image Computing and Computer-Assisted Intervention, pages 375–385, 2022

    work page 2022

  35. [35]

    An enhanced multi- modal brain graph network for classifying neuropsychiatric disorders.Medical Image Analysis, 81:102550, 2022

    Liangliang Liu, Yu-Ping Wang, Yi Wang, Pei Zhang, and Shufeng Xiong. An enhanced multi- modal brain graph network for classifying neuropsychiatric disorders.Medical Image Analysis, 81:102550, 2022

    work page 2022

  36. [36]

    Graph neural network for interpreting task-fMRI biomarkers

    Xiaoxiao Li, Nicha C Dvornek, Yuan Zhou, Juntang Zhuang, Pamela Ventola, and James S Duncan. Graph neural network for interpreting task-fMRI biomarkers. InInternational Conference on Medical Image Computing and Computer-Assisted Intervention, pages 485–493, 2019

    work page 2019

  37. [37]

    Tensor graph convolutional neural network

    Tong Zhang, Wenming Zheng, Zhen Cui, and Yang Li. Tensor graph convolutional neural network, 2018. URLhttps://arxiv.org/abs/1803.10071

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  38. [38]

    Tensor graph neural networks for learning on time varying graphs

    Osman Asif Malik, Shashanka Ubaru, Lior Horesh, Misha E Kilmer, and Haim Avron. Tensor graph neural networks for learning on time varying graphs. InProceedings of NIPS Workshop, 2019

    work page 2019

  39. [39]

    Mr-gcn: Multi-relational graph convolutional networks based on generalized tensor product

    Zhichao Huang, Xutao Li, Yunming Ye, and Michael K Ng. Mr-gcn: Multi-relational graph convolutional networks based on generalized tensor product. InIJCAI, volume 20, pages 1258–1264, 2020

    work page 2020

  40. [40]

    ALERT: Atlas-based low estimation rank tensor approach to detect Autism spectrum disorder

    Ananya Samanta, Monalisa Sarma, and Debasis Samanta. ALERT: Atlas-based low estimation rank tensor approach to detect Autism spectrum disorder. InInternational Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), pages 1–4, 2023

    work page 2023

  41. [41]

    Classification study of Alzheimer’s disease based on self-attention mechanism and dti imaging using gcn.IEEE Access, 12:24387–24395, 2024

    Yilin Sang and Wan Li. Classification study of Alzheimer’s disease based on self-attention mechanism and dti imaging using gcn.IEEE Access, 12:24387–24395, 2024. 12

    work page 2024

  42. [42]

    Brain structure- function fusing representation learning using adversarial decomposed-V AE for analyzing MCI

    Qiankun Zuo, Ning Zhong, Yi Pan, Huisi Wu, Baiying Lei, and Shuqiang Wang. Brain structure- function fusing representation learning using adversarial decomposed-V AE for analyzing MCI. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 31:4017–4028, 2023

    work page 2023

  43. [43]

    Tensor-based complex- valued graph neural network for dynamic coupling multimodal brain networks

    Yanwu Yang, Guoqing Cai, Chenfei Ye, Yang Xiang, and Ting Ma. Tensor-based complex- valued graph neural network for dynamic coupling multimodal brain networks. InIEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2023

    work page 2023

  44. [44]

    TGNet: tensor-based graph convolutional networks for multimodal brain network analysis.BioData Mining, 17(1):55, 2024

    Zhaoming Kong, Rong Zhou, Xinwei Luo, Songlin Zhao, Ann B Ragin, Alex D Leow, and Lifang He. TGNet: tensor-based graph convolutional networks for multimodal brain network analysis.BioData Mining, 17(1):55, 2024

    work page 2024

  45. [45]

    Community detection in graphs.Physics Reports, 486(3-5):75–174, 2010

    Santo Fortunato. Community detection in graphs.Physics Reports, 486(3-5):75–174, 2010

    work page 2010

  46. [46]

    Encoder-decoder symmetric nonnegative matrix tri- factorization for graph clustering

    Amjad Seyedi and Nicolas Gillis. Encoder-decoder symmetric nonnegative matrix tri- factorization for graph clustering. InIEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 501–505, 2026

    work page 2026

  47. [47]

    Stochastic blockmodels and community structure in networks.Physical Review E, 83(1):016107, 2011

    Brian Karrer and Mark EJ Newman. Stochastic blockmodels and community structure in networks.Physical Review E, 83(1):016107, 2011

    work page 2011

  48. [48]

    Mixed membership stochastic blockmodels.Journal of Machine Learning Research, 9(Sep):1981–2014, 2008

    Edoardo M Airoldi, David M Blei, Stephen E Fienberg, and Eric P Xing. Mixed membership stochastic blockmodels.Journal of Machine Learning Research, 9(Sep):1981–2014, 2008

    work page 1981

  49. [49]

    Deep asymmetric nonnegative matrix factorization for graph clustering.Pattern Recognition, 148:110179, 2024

    Akram Hajiveiseh, Seyed Amjad Seyedi, and Fardin Akhlaghian Tab. Deep asymmetric nonnegative matrix factorization for graph clustering.Pattern Recognition, 148:110179, 2024

    work page 2024

  50. [50]

    A deep matrix factorization method for learning attribute representations.IEEE Transactions on Pattern Analysis and Machine Intelligence, 39(3):417–429, 2016

    George Trigeorgis, Konstantinos Bousmalis, Stefanos Zafeiriou, and Björn W Schuller. A deep matrix factorization method for learning attribute representations.IEEE Transactions on Pattern Analysis and Machine Intelligence, 39(3):417–429, 2016

    work page 2016

  51. [51]

    Learning the parts of objects by non-negative matrix factorization.Nature, 401(6755):788–791, 1999

    Daniel D Lee and H Sebastian Seung. Learning the parts of objects by non-negative matrix factorization.Nature, 401(6755):788–791, 1999

    work page 1999

  52. [52]

    Exponentially convergent algorithms for supervised matrix factorization

    Joowon Lee, Hanbaek Lyu, and Weixin Yao. Exponentially convergent algorithms for supervised matrix factorization. InAdvances in Neural Information Processing Systems, volume 36, pages 76947–76959, 2023

    work page 2023

  53. [53]

    Supervised matrix factorization: local landscape analysis and applications

    Joowon Lee, Hanbaek Lyu, and Weixin Yao. Supervised matrix factorization: local landscape analysis and applications. InInternational Conference on Machine Learning, 2024

    work page 2024

  54. [54]

    Implicit regularization in deep matrix factorization

    Sanjeev Arora, Nadav Cohen, Wei Hu, and Yuping Luo. Implicit regularization in deep matrix factorization. InAdvances in Neural Information Processing Systems, volume 32, 2019

    work page 2019

  55. [55]

    A consistent and flexible framework for deep matrix factorizations.Pattern Recognition, 134:109102, 2023

    Pierre De Handschutter and Nicolas Gillis. A consistent and flexible framework for deep matrix factorizations.Pattern Recognition, 134:109102, 2023

    work page 2023

  56. [56]

    A convergence theory for deep learning via over-parameterization

    Zeyuan Allen-Zhu, Yuanzhi Li, and Zhao Song. A convergence theory for deep learning via over-parameterization. InInternational Conference on Machine Learning, pages 242–252, 2019

    work page 2019

  57. [57]

    Correspondence of the brain’s functional architecture during activation and rest.Proceedings of the National Academy of Sciences, 106(31):13040–13045, 2009

    Stephen M Smith, Peter T Fox, Karla L Miller, David C Glahn, P Mickle Fox, Clare E Mackay, Nicola Filippini, Kate E Watkins, Roberto Toro, Angela R Laird, et al. Correspondence of the brain’s functional architecture during activation and rest.Proceedings of the National Academy of Sciences, 106(31):13040–13045, 2009

    work page 2009

  58. [58]

    The organization of the human cerebral cortex estimated by intrinsic functional connectivity

    BT Thomas Yeo, Fenna M Krienen, Jorge Sepulcre, Mert R Sabuncu, Danial Lashkari, Marisa Hollinshead, Joshua L Roffman, Jordan W Smoller, Lilla Zöllei, Jonathan R Polimeni, et al. The organization of the human cerebral cortex estimated by intrinsic functional connectivity. Journal of Neurophysiology, 2011

    work page 2011

  59. [59]

    Different roles of frontal versus striatal atrophy in HIV-associated neurocognitive disorders.Human Brain Mapping, 40(10):3010–3026, 2019

    Sarah M Israel, Shiva Hassanzadeh-Behbahani, Peter E Turkeltaub, David J Moore, Ronald J Ellis, and Xiong Jiang. Different roles of frontal versus striatal atrophy in HIV-associated neurocognitive disorders.Human Brain Mapping, 40(10):3010–3026, 2019. 13

    work page 2019

  60. [60]

    Novel neuroimaging methods to understand how HIV affects the brain.Current HIV/AIDS Reports, 12(2):289–298, 2015

    Paul M Thompson and Neda Jahanshad. Novel neuroimaging methods to understand how HIV affects the brain.Current HIV/AIDS Reports, 12(2):289–298, 2015

    work page 2015

  61. [61]

    Modular brain networks.Annual Review of Psychology, 67 (1):613–640, 2016

    Olaf Sporns and Richard F Betzel. Modular brain networks.Annual Review of Psychology, 67 (1):613–640, 2016

    work page 2016

  62. [62]

    Age exacerbates HIV- associated white matter abnormalities.Journal of Neurovirology, 22(2):201–212, 2016

    Talia R Seider, Assawin Gongvatana, Adam J Woods, Huaihou Chen, Eric C Porges, Tiffany Cummings, Stephen Correia, Karen Tashima, and Ronald A Cohen. Age exacerbates HIV- associated white matter abnormalities.Journal of Neurovirology, 22(2):201–212, 2016

    work page 2016

  63. [63]

    Spontaneous fluctuations in brain activity observed with functional magnetic resonance imaging.Nature Reviews Neuroscience, 8(9):700–711, 2007

    Michael D Fox and Marcus E Raichle. Spontaneous fluctuations in brain activity observed with functional magnetic resonance imaging.Nature Reviews Neuroscience, 8(9):700–711, 2007

    work page 2007

  64. [64]

    Functional integration between the two brain hemispheres: evidence from the homotopic functional connectivity under resting state.Frontiers in Neuroscience, 14:932, 2020

    Xinhu Jin, Xinyu Liang, and Gaolang Gong. Functional integration between the two brain hemispheres: evidence from the homotopic functional connectivity under resting state.Frontiers in Neuroscience, 14:932, 2020

    work page 2020

  65. [65]

    The influence of mild carbon dioxide on brain functional homotopy using resting-state fMRI

    Olga Marshall, Jinsoo Uh, Daniel Lurie, Hanzhang Lu, Michael P Milham, and Yulin Ge. The influence of mild carbon dioxide on brain functional homotopy using resting-state fMRI. Human Brain Mapping, 36(10):3912–3921, 2015

    work page 2015

  66. [66]

    Brainnet viewer: a network visualization tool for human brain connectomics.PloS One, 8(7):e68910, 2013

    Mingrui Xia, Jinhui Wang, and Yong He. Brainnet viewer: a network visualization tool for human brain connectomics.PloS One, 8(7):e68910, 2013

    work page 2013

  67. [67]

    A meta-analysis of gray matter volume abnormalities in HIV patients.Psychiatry Research: Neuroimaging, 335:111722, 2023

    Mingxiang Xu and Xing-Da Ju. A meta-analysis of gray matter volume abnormalities in HIV patients.Psychiatry Research: Neuroimaging, 335:111722, 2023

    work page 2023

  68. [68]

    Divergent structural and functional brain alterations in HIV-infected patients: a multimodal meta-analysis.Frontiers in Neurology, 16:1618408, 2025

    Zhong Li, Xingxing Jin, Meng Zhang, Hongxia Wang, Wangyi Liu, Beiran Wang, Baolin Wu, and Xuekun Li. Divergent structural and functional brain alterations in HIV-infected patients: a multimodal meta-analysis.Frontiers in Neurology, 16:1618408, 2025

    work page 2025

  69. [69]

    Hou, and Max Tegmark

    Ziming Liu, Yixuan Wang, Sachin Vaidya, Fabian Ruehle, James Halverson, Marin Soljacic, Thomas Y . Hou, and Max Tegmark. KAN: Kolmogorov–Arnold networks. InInternational Conference on Learning Representations, 2025

    work page 2025

  70. [70]

    SIAM, 2020

    Nicolas Gillis.Nonnegative Matrix Factorization. SIAM, 2020

    work page 2020

  71. [71]

    Learning topic models–provably and efficiently.Communications of the ACM, 61(4):85–93, 2018

    Sanjeev Arora, Rong Ge, Yoni Halpern, David Mimno, Ankur Moitra, David Sontag, Yichen Wu, and Michael Zhu. Learning topic models–provably and efficiently.Communications of the ACM, 61(4):85–93, 2018

    work page 2018

  72. [72]

    Anchor-free correlated topic modeling: Identifiability and algorithm

    Kejun Huang, Xiao Fu, and Nikolaos D Sidiropoulos. Anchor-free correlated topic modeling: Identifiability and algorithm. InAdvances in Neural Information Processing Systems, volume 29, 2016

    work page 2016

  73. [73]

    Anchor-free correlated topic modeling.IEEE transactions on Pattern Analysis and Machine Intelligence, 41 (5):1056–1071, 2018

    Xiao Fu, Kejun Huang, Nicholas D Sidiropoulos, Qingjiang Shi, and Mingyi Hong. Anchor-free correlated topic modeling.IEEE transactions on Pattern Analysis and Machine Intelligence, 41 (5):1056–1071, 2018

    work page 2018

  74. [74]

    Ragin, Hongyan Du, Renee Ochs, Ying Wu, Christina L

    Ann B. Ragin, Hongyan Du, Renee Ochs, Ying Wu, Christina L. Sammet, Alfred Shoukry, and Leon G. Epstein. Structural brain alterations can be detected early in hiv infection.Neurology, 79(24):2328–2334, 2012

    work page 2012

  75. [75]

    Identifying HIV-induced subgraph patterns in brain networks with side information.Brain Informatics, 2 (4):211–223, 2015

    Bokai Cao, Xiangnan Kong, Jingyuan Zhang, Philip S Yu, and Ann B Ragin. Identifying HIV-induced subgraph patterns in brain networks with side information.Brain Informatics, 2 (4):211–223, 2015

    work page 2015

  76. [76]

    Altshuler

    Olusola Ajilore, Nathalie Vizueta, Patricia Walshaw, Liang Zhan, Alex Leow, and Lori L. Altshuler. Connectome signatures of neurocognitive abnormalities in euthymic bipolar I disorder.Journal of Psychiatric Research, 68:37–44, 2015

    work page 2015

  77. [77]

    Conn: a functional connectivity toolbox for correlated and anticorrelated brain networks.Brain Connectivity, 2(3):125–141, 2012

    Susan Whitfield-Gabrieli and Alfonso Nieto-Castanon. Conn: a functional connectivity toolbox for correlated and anticorrelated brain networks.Brain Connectivity, 2(3):125–141, 2012. 14

    work page 2012

  78. [78]

    stretches

    Liang Zhan, Jiayu Zhou, Yalin Wang, Yan Jin, Neda Jahanshad, Gautam Prasad, Talia M Nir, Cassandra D Leonardo, Jieping Ye, Paul M Thompson, et al. Comparison of nine tractography algorithms for detecting abnormal structural brain networks in Alzheimer’s disease.Frontiers in Aging Neuroscience, 7:48, 2015. Appendix A Optimization model In this section, we ...

    work page 2015