pith. sign in

arxiv: 2605.18316 · v1 · pith:M3IRBZLYnew · submitted 2026-05-18 · 💻 cs.LG · cs.GR

Dynamic Elliptical Graph Factor Models via Riemannian Optimization with Geodesic Temporal Regularization

Chuansen Peng , Xiaojing Shen This is my paper

Pith reviewed 2026-05-20 12:13 UTC · model grok-4.3

classification 💻 cs.LG cs.GR
keywords dynamic graph estimationRiemannian optimizationGrassmann manifoldprecision matrixfactor modelstemporal regularizationgeodesic penalty
0
0 comments X

The pith

A factor model on the Grassmann manifold with geodesic penalties estimates time-varying precision matrices more reliably in small-sample settings.

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

The paper develops an algorithm for recovering sequences of graphs from high-dimensional observations collected over time. It represents each precision matrix as a low-rank-plus-diagonal object controlled by a latent elliptical graph factor model, which cuts the number of free parameters enough to make estimation practical when observations per window are few. Smoothness across time is imposed by penalizing geodesic distances on the Grassmann manifold rather than Euclidean distances, and the resulting problem is solved by a Riemannian gradient-descent procedure that stays on the manifold at every step. Experiments on synthetic cases and real datasets show consistent gains over earlier methods in recovering the underlying graph trajectory.

Core claim

The authors claim that a latent elliptical graph factor model imposes a low-rank-plus-diagonal structure on time-varying precision matrices, and that adding a geodesic penalty on the Grassmann manifold produces temporally coherent estimates; an efficient Riemannian gradient-descent solver then recovers the sequence and converges to a stationary point.

What carries the argument

The DEGfM procedure, which combines a latent elliptical graph factor model with Riemannian optimization on the Grassmann manifold and geodesic temporal regularization.

If this is right

  • Reliable graph recovery becomes possible even when the number of variables greatly exceeds the number of samples in each time window.
  • The recovered graph sequence changes smoothly along the intrinsic geometry of the manifold instead of exhibiting artificial jumps in ambient Euclidean space.
  • The optimization procedure is guaranteed to reach a stationary point of the non-convex objective.
  • Performance exceeds that of prior Euclidean and non-factor methods on both synthetic benchmarks and real data.

Where Pith is reading between the lines

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

  • The same geodesic-regularization idea could be applied to other manifold-valued time series such as covariance trajectories in finance.
  • Replacing the elliptical factor model with a different low-dimensional parametrization might extend the approach to graphs that are not well approximated by low-rank-plus-diagonal forms.
  • The method supplies a concrete way to test whether temporal coherence in real networks is better captured by manifold geodesics than by Euclidean penalties.

Load-bearing premise

The sequence of precision matrices admits a low-rank-plus-diagonal decomposition controlled by a latent elliptical graph factor model.

What would settle it

A collection of high-dimensional time series whose true precision matrices lack the low-rank-plus-diagonal structure, followed by a check that DEGfM loses its reported advantage over Euclidean baselines.

Figures

Figures reproduced from arXiv: 2605.18316 by Chuansen Peng, Xiaojing Shen.

Figure 1
Figure 1. Figure 1: Illustration of the key geometric operations in Riemannian optimiza [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: ROC curves for graph recovery in the non-LRaD setting under four [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Mean AUC comparison in the LRaD setting for Gaussian and Student [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Running time versus the number of nodes p on a logarithmic scale. The results show that the proposed method scales favorably and remains computationally efficient as the graph dimension increases. Θt = YtY⊤ t + Dt, where Yt ∈ R 50×10 is a sparse factor loading matrix with non-zero entries drawn from U(1, 3) and sparsified to retain approximately 20% of entries, and Dt is a diagonal matrix with positive dia… view at source ↗
Figure 6
Figure 6. Figure 6: Graph modularity Q estimated by each algorithm across T = 64 quarterly windows on the S&P 500 dataset (2005–2020). Shaded regions mark the three major market stress episodes: the 2008 Global Financial Crisis, the European Sovereign Debt Crisis (2010–2012), and the COVID-19 pandemic shock (2020). Vertical dashed lines indicate each crisis peak. DEGFM (solid blue) attains the highest modularity during tranqu… view at source ↗
Figure 7
Figure 7. Figure 7: Error bars denote one standard deviation across five-fold time-series [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: Sector-averaged partial correlation matrices ( [PITH_FULL_IMAGE:figures/full_fig_p012_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Multi-metric performance radar chart comparing the four algorithms [PITH_FULL_IMAGE:figures/full_fig_p013_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Riemannian gradient norm ∥grad F∥ versus iteration count for each algorithm on a representative quarterly window of the S&P 500 dataset, displayed on linear (left) and logarithmic (right) scales. The horizontal dotted line marks the convergence tolerance. ces onto a three-dimensional principal component space and examine how the resulting latent embeddings evolve across three historically significant mark… view at source ↗
Figure 13
Figure 13. Figure 13: Quantitative performance comparison of DEGFM, IndGGFM, [PITH_FULL_IMAGE:figures/full_fig_p014_13.png] view at source ↗
Figure 15
Figure 15. Figure 15: Resting-state network (RSN) community detection results on the [PITH_FULL_IMAGE:figures/full_fig_p015_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Convergence analysis and hyperparameter sensitivity of DEGFM [PITH_FULL_IMAGE:figures/full_fig_p015_16.png] view at source ↗
read the original abstract

Inferring time-varying graph structures from high-dimensional nodal observations is a fundamental problem arising in neuroscience, finance, climatology, and beyond. Two intrinsic challenges govern this problem: maintaining the \emph{temporal coherence} of the latent graph across successive observation windows, and respecting the \emph{intrinsic Riemannian geometry} of the symmetric positive definite manifold on which precision matrices naturally reside, a curved space whose geodesic structure departs fundamentally from that of the ambient Euclidean space. In this paper we propose dynamic estimation on the Grassmann manifold with a factor model (\textsc{Degfm}), a novel algorithm that jointly addresses both challenges. We model the time-varying precision matrix sequence as a low-rank-plus-diagonal structure governed by a latent elliptical graph factor model, which drastically reduces the effective parameter count and enables reliable estimation in the challenging small-sample regime. Temporal coherence is enforced through a Riemannian geodesic penalty defined on the Grassmann manifold, ensuring that the estimated graph trajectory is smooth with respect to the intrinsic geometry rather than the ambient Euclidean space. To solve the resulting non-convex optimization problem over Grassmann-manifold-valued sequences subject to the LRaD constraint, we derive an efficient Riemannian gradient descent algorithm that respects the manifold structure at every iterate and rigorously establish its convergence to a stationary point. Extensive experiments on both synthetic benchmarks and real-world datasets demonstrate that \textsc{Degfm} consistently outperforms state-of-the-art baselines across all evaluation metrics, confirming the practical effectiveness of the proposed framework.

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 / 1 minor

Summary. The paper claims to introduce DEGfM, a method for dynamic estimation of time-varying graph structures by modeling precision matrix sequences as low-rank-plus-diagonal structures under a latent elliptical graph factor model on the Grassmann manifold. It uses a Riemannian geodesic penalty for temporal coherence and derives a Riemannian gradient descent algorithm with convergence to a stationary point, demonstrating outperformance on synthetic and real-world datasets.

Significance. Should the low-rank-plus-diagonal assumption be appropriate for the data, this work could advance the field by providing a parameter-efficient, geometry-aware approach to small-sample dynamic precision matrix estimation. The combination of factor models with manifold optimization and the convergence guarantee are notable strengths that could influence future research in time-series graphical models.

major comments (2)
  1. [Abstract] The headline claim of reliable small-sample estimation rests on the time-varying precision matrices following a low-rank-plus-diagonal structure. The abstract asserts outperformance on real datasets but supplies no diagnostic such as singular-value decay or reconstruction error under the LRaD model to confirm the structure holds on those datasets. This is a load-bearing assumption for the central claim.
  2. [Algorithm and Convergence Analysis] The derivation of the efficient Riemannian gradient descent algorithm and the rigorous proof of convergence to a stationary point are key contributions. The manuscript should provide more explicit details on how the geodesic temporal regularization is incorporated into the Riemannian updates to allow full verification of the technical claims.
minor comments (1)
  1. The notation for the Grassmann manifold and the factor model parameters could be clarified with a dedicated notation table for reader convenience.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major comment below and describe the revisions we will make to strengthen the presentation and verifiability of our results.

read point-by-point responses

  1. Referee: [Abstract] The headline claim of reliable small-sample estimation rests on the time-varying precision matrices following a low-rank-plus-diagonal structure. The abstract asserts outperformance on real datasets but supplies no diagnostic such as singular-value decay or reconstruction error under the LRaD model to confirm the structure holds on those datasets. This is a load-bearing assumption for the central claim.

    Authors: We agree that empirical verification of the low-rank-plus-diagonal (LRaD) structure on the real-world datasets would strengthen the central claim. In the revised manuscript we will add diagnostics, including singular-value decay plots and reconstruction-error curves under the LRaD model, for each real dataset. These will be placed in a new subsection of the experimental results and referenced from the abstract to make the load-bearing assumption explicit and testable. revision: yes

  2. Referee: [Algorithm and Convergence Analysis] The derivation of the efficient Riemannian gradient descent algorithm and the rigorous proof of convergence to a stationary point are key contributions. The manuscript should provide more explicit details on how the geodesic temporal regularization is incorporated into the Riemannian updates to allow full verification of the technical claims.

    Authors: We appreciate the request for greater algorithmic transparency. In the revision we will expand the algorithm section with an explicit derivation of the Riemannian gradient of the geodesic temporal regularization term, including the closed-form expression for its projection onto the tangent space of the Grassmann manifold at each iterate. We will also supply updated pseudocode that isolates the regularization contribution within the overall update rule, thereby allowing direct verification of the convergence analysis. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected in derivation chain

full rationale

The paper states an explicit modeling assumption that time-varying precision matrices admit a low-rank-plus-diagonal structure under a latent elliptical graph factor model, then constructs a Riemannian gradient descent procedure on the Grassmann manifold equipped with a geodesic temporal penalty and proves convergence to a stationary point. None of these steps reduce by the paper's own equations to quantities defined in terms of the fitted outputs or to self-citations that themselves presuppose the target result. The algorithm derivation follows standard manifold optimization techniques applied to the stated constraint, and the performance claims are presented as empirical outcomes rather than identities forced by the modeling choice. No self-definitional, fitted-input-renamed-as-prediction, or load-bearing self-citation patterns appear in the abstract or described derivation.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim rests on the domain assumption that precision matrices inhabit a Riemannian manifold whose geodesic structure must be respected, plus the modeling choice of a low-rank-plus-diagonal elliptical factor structure whose validity is not independently verified in the abstract.

axioms (1)
  • domain assumption Precision matrices reside on the symmetric positive definite manifold whose geodesic structure departs fundamentally from Euclidean space.
    Stated explicitly as one of the two intrinsic challenges governing the problem.
invented entities (1)
  • Dynamic Elliptical Graph Factor Model (DEGfM) no independent evidence
    purpose: To represent time-varying precision matrices via low-rank-plus-diagonal structure and enforce temporal coherence through geodesic penalty.
    Newly introduced framework whose effectiveness is asserted via the optimization procedure and experiments.

pith-pipeline@v0.9.0 · 5796 in / 1457 out tokens · 87862 ms · 2026-05-20T12:13:36.692480+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

89 extracted references · 89 canonical work pages · 2 internal anchors

  1. [1]

    The structure and function of complex networks

    Mark E J Newman. The structure and function of complex networks. SIAM Review, 45(2):167–256, 2003

    work page 2003

  2. [2]

    Collec- tive dynamics of ‘small-world’ networks

    Duncan J Watts and Steven H Strogatz. Collec- tive dynamics of ‘small-world’ networks. Nature, 393(6684):440–442, 1998

    work page 1998

  3. [3]

    Emergence of scaling in random networks

    Albert-L ´aszl´o Barab ´asi and R ´eka Albert. Emergence of scaling in random networks. Science, 286(5439):509– 512, 1999

    work page 1999

  4. [4]

    Statistical analysis of network data with R, 2014

    Eric D Kolaczyk. Statistical analysis of network data with R, 2014

    work page 2014

  5. [5]

    The emerging field of signal processing on graphs: Extending high- dimensional data analysis to networks and other irregular domains

    David I Shuman, Sunil K Narang, Pascal Frossard, An- tonio Ortega, and Pierre Vandergheynst. The emerging field of signal processing on graphs: Extending high- dimensional data analysis to networks and other irregular domains. IEEE Signal Processing Magazine , 30(3):83– 98, 2013

    work page 2013

  6. [6]

    Graph signal processing: Overview, challenges, and applications

    Antonio Ortega, Pascal Frossard, Jelena Kova ˇcevi´c, Jos´e M F Moura, and Pierre Vandergheynst. Graph signal processing: Overview, challenges, and applications. Pro- ceedings of the IEEE , 106(5):808–828, 2018

    work page 2018

  7. [7]

    Learning graphs from data: A signal representation perspective

    Xiaowen Dong, Dorina Thanou, Michael Rabbat, and Pascal Frossard. Learning graphs from data: A signal representation perspective. IEEE Signal Processing Magazine, 36(3):44–63, 2019

    work page 2019

  8. [8]

    Connecting the dots: Identifying network structure via graph signal processing

    Gonzalo Mateos, Santiago Segarra, Antonio G Marques, and Alejandro Ribeiro. Connecting the dots: Identifying network structure via graph signal processing. IEEE Signal Processing Magazine , 36(3):16–43, 2019

    work page 2019

  9. [9]

    Inductive representation learning on large graphs

    William L Hamilton, Rex Ying, and Jure Leskovec. Inductive representation learning on large graphs. In Advances in Neural Information Processing Sys- tems (NeurIPS), volume 30, 2017

    work page 2017

  10. [10]

    Convolutional neural networks on graphs with fast localized spectral filtering

    Micha ¨el Defferrard, Xavier Bresson, and Pierre Van- dergheynst. Convolutional neural networks on graphs with fast localized spectral filtering. In Advances in Neural Information Processing Systems (NeurIPS) , vol- ume 29, 2016

    work page 2016

  11. [11]

    Semi-supervised clas- sification with graph convolutional networks

    Thomas N Kipf and Max Welling. Semi-supervised clas- sification with graph convolutional networks. In Interna- tional Conference on Learning Representations (ICLR) , 2017

    work page 2017

  12. [12]

    Spectral networks and locally connected networks on graphs

    Joan Bruna, Wojciech Zaremba, Arthur Szlam, and Yann LeCun. Spectral networks and locally connected networks on graphs. In International Conference on Learning Representations (ICLR) , 2014

    work page 2014

  13. [13]

    Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gauges

    Michael M Bronstein, Joan Bruna, Taco Cohen, and Petar Veli ˇckovi´c. Geometric deep learning: Grids, groups, graphs, geodesics, and gauges. arXiv preprint arXiv:2104.13478, 2021

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  14. [14]

    Graph neural networks: foundation, frontiers and applications

    Lingfei Wu, Peng Cui, Jian Pei, Liang Zhao, and Xiaojie Guo. Graph neural networks: foundation, frontiers and applications. In Proceedings of the 28th ACM SIGKDD conference on knowledge discovery and data mining , pages 4840–4841, 2022

    work page 2022

  15. [15]

    Deep Learning on Graphs

    Yao Ma and Jiliang Tang. Deep Learning on Graphs . Cambridge University Press, Cambridge, 2021

    work page 2021

  16. [16]

    Estimating time-varying brain connectivity networks from fMRI time series

    Ricardo P Monti, Peter Hellyer, David Sharp, Robert Leech, Christoforos Anagnostopoulos, and Giovanni Montana. Estimating time-varying brain connectivity networks from fMRI time series. NeuroImage, 103:427– 443, 2014

    work page 2014

  17. [17]

    Hierarchical structure in financial markets

    Rosario N Mantegna. Hierarchical structure in financial markets. European Physical Journal B , 11(1):193–197, 1999

    work page 1999

  18. [18]

    Dynamics of market correlations: Taxonomy and portfolio analysis

    Jukka-Pekka Onnela, Anirban Chakraborti, Kimmo 25 Kaski, J ´anos Kert ´esz, and Antti Kanto. Dynamics of market correlations: Taxonomy and portfolio analysis. Physical Review E , 68(5):056110, 2003

    work page 2003

  19. [19]

    Estimating time-varying networks

    Mladen Kolar, Le Song, Amr Ahmed, and Eric P Xing. Estimating time-varying networks. Annals of Applied Statistics, 4(1):94–123, 2010

    work page 2010

  20. [20]

    Time varying undirected graphs

    Shuheng Zhou, John Lafferty, and Larry Wasserman. Time varying undirected graphs. Machine Learning , 80(2–3):295–319, 2010

    work page 2010

  21. [21]

    Recovering time-varying networks of dependencies in social and biological stud- ies

    Amr Ahmed and Eric P Xing. Recovering time-varying networks of dependencies in social and biological stud- ies. Proceedings of the National Academy of Sciences , 106(29):11878–11883, 2009

    work page 2009

  22. [22]

    Network inference via the time-varying graph- ical Lasso

    David Hallac, Youngsuk Park, Stephen Boyd, and Jure Leskovec. Network inference via the time-varying graph- ical Lasso. In ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD), pages 205–213, 2017

    work page 2017

  23. [23]

    Modeling changing dependency structure in multivariate time series

    Xing Xuan and Kevin Murphy. Modeling changing dependency structure in multivariate time series. In International Conference on Machine Learning (ICML) , pages 1055–1062, 2007

    work page 2007

  24. [24]

    Representation learning for dynamic graphs: A survey

    Seyed Mehran Kazemi, Rishab Goel, Kshitij Jain, Ivan Kobyzev, Akshay Sethi, Peter Forsyth, and Pascal Poupart. Representation learning for dynamic graphs: A survey. Journal of Machine Learning Research , 21(70):1–73, 2020

    work page 2020

  25. [25]

    Foundations and modeling of dynamic networks using dynamic graph neural networks: A survey

    Joakim Skarding, Bogdan Gabrys, and Katarzyna Mu- sial. Foundations and modeling of dynamic networks using dynamic graph neural networks: A survey. IEEE Access, 9:79143–79168, 2021

    work page 2021

  26. [26]

    Statistical Analysis of Network Data: Methods and Models

    Eric D Kolaczyk. Statistical Analysis of Network Data: Methods and Models . Springer, New York, 2009

    work page 2009

  27. [27]

    Graphical Models

    Steffen L Lauritzen. Graphical Models. Oxford Univer- sity Press, Oxford, 1996

    work page 1996

  28. [28]

    Learning Laplacian matrix in smooth graph signal representations

    Xiaowen Dong, Dorina Thanou, Pascal Frossard, and Pierre Vandergheynst. Learning Laplacian matrix in smooth graph signal representations. IEEE Transactions on Signal Processing , 64(23):6160–6173, 2016

    work page 2016

  29. [29]

    Graph learning from data under Laplacian and structural constraints

    Hilmi E Egilmez, Eduardo Pavez, and Antonio Ortega. Graph learning from data under Laplacian and structural constraints. IEEE Journal of Selected Topics in Signal Processing, 11(6):811–825, 2017

    work page 2017

  30. [30]

    Characteri- zation and inference of graph diffusion processes from observations of stationary signals

    Bastien Pasdeloup, Vincent Gripon, Gr ´egoire Mercier, Dominique Pastor, and Michael G Rabbat. Characteri- zation and inference of graph diffusion processes from observations of stationary signals. IEEE Transactions on Signal and Information Processing over Networks , 4(3):481–496, 2018

    work page 2018

  31. [31]

    Joint network topology inference in the presence of hidden nodes

    Madeline Navarro, Samuel Rey, Andrei Buciulea, Anto- nio G Marques, and Santiago Segarra. Joint network topology inference in the presence of hidden nodes. IEEE Transactions on Signal Processing, 72:2710–2725, 2024

    work page 2024

  32. [32]

    Deep graph structure learning for robust representations: A survey

    Yanqiao Zhu, Weizhi Xu, Jinghao Zhang, Yuanqi Du, Jieyu Zhang, Qiang Liu, Carl Yang, and Shu Wu. A survey on graph structure learning: Progress and oppor- tunities. arXiv preprint arXiv:2103.03036 , 2021

    work page arXiv 2021

  33. [33]

    Covariance selection

    Arthur P Dempster. Covariance selection. Biometrics, 28(1):157–175, 1972

    work page 1972

  34. [34]

    Friedman, T

    J. Friedman, T. Hastie, and R. Tibshirani. Sparse inverse covariance estimation with the graphical Lasso. Biostatistics, 9(3):432–441, 2008

    work page 2008

  35. [35]

    Model selection through sparse maximum likelihood estimation for multivariate Gaussian or binary data

    Onureena Banerjee, Laurent El Ghaoui, and Alexandre d’Aspremont. Model selection through sparse maximum likelihood estimation for multivariate Gaussian or binary data. Journal of Machine Learning Research, 9:485–516, 2008

    work page 2008

  36. [36]

    High- dimensional graphs and variable selection with the Lasso

    Nicolai Meinshausen and Peter B ¨uhlmann. High- dimensional graphs and variable selection with the Lasso. Annals of Statistics , 34(3):1436–1462, 2006

    work page 2006

  37. [37]

    Model selection and estimation in the Gaussian graphical model

    Ming Yuan and Yi Lin. Model selection and estimation in the Gaussian graphical model. Biometrika, 94(1):19– 35, 2007

    work page 2007

  38. [38]

    Regression shrinkage and selection via the Lasso

    Robert Tibshirani. Regression shrinkage and selection via the Lasso. Journal of the Royal Statistical Society, Series B, 58(1):267–288, 1996

    work page 1996

  39. [39]

    Latent variable graphical model selection via convex optimization

    Venkat Chandrasekaran, Pablo A Parrilo, and Alan S Willsky. Latent variable graphical model selection via convex optimization. Annals of Statistics , 40(4):1935– 2013, 2012

    work page 1935

  40. [40]

    How to learn a graph from smooth signals

    Vassilis Kalofolias. How to learn a graph from smooth signals. In International Conference on Artificial Intel- ligence and Statistics (AISTATS) , pages 920–929, 2016

    work page 2016

  41. [41]

    Network topology inference from spectral templates

    Santiago Segarra, Antonio G Marques, Gonzalo Mateos, and Alejandro Ribeiro. Network topology inference from spectral templates. IEEE Transactions on Signal and Information Processing over Networks , 3(3):467–483, 2017

    work page 2017

  42. [42]

    Kumar, J

    S. Kumar, J. Ying, J. V . de M. Cardoso, and D. P. Palomar. A unified framework for structured graph learning via spectral constraints. Journal of Machine Learning Research, 21(22):1–60, 2020

    work page 2020

  43. [43]

    Does the ℓ1-norm learn a sparse graph under Laplacian constrained graphical models? IEEE Transactions on Signal Processing, 70:4685–4698, 2022

    Jiaxi Ying, Jos ´e Vin ´ıcius de Miranda Cardoso, and Daniel P Palomar. Does the ℓ1-norm learn a sparse graph under Laplacian constrained graphical models? IEEE Transactions on Signal Processing, 70:4685–4698, 2022

    work page 2022

  44. [44]

    Learning to learn graph topolo- gies

    Wenqian Pu, Yang Cao, Hanghang Liu, Caihua Shen, Jie Chen, and Jieping Ye. Learning to learn graph topolo- gies. In Advances in Neural Information Processing Systems (NeurIPS), volume 34, pages 4280–4293, 2021

    work page 2021

  45. [45]

    SLAPS: Self-supervision improves structure learning for graph neural networks

    Bahare Fatemi, Layla El Asri, and Seyed Mehran Kazemi. SLAPS: Self-supervision improves structure learning for graph neural networks. In Advances in Neural Information Processing Systems (NeurIPS) , vol- ume 34, pages 22667–22681, 2021

    work page 2021

  46. [46]

    Acceler- ated graph learning from smooth signals

    Seyed Saman Saboksayr and Gonzalo Mateos. Acceler- ated graph learning from smooth signals. IEEE Signal Processing Letters, 28:2192–2196, 2021

    work page 2021

  47. [47]

    Nonconvex sparse graph learning un- der Laplacian constrained graphical model

    Jos ´e Vin ´ıcius de Miranda Cardoso, Jiaxi Ying, and Daniel P Palomar. Nonconvex sparse graph learning un- der Laplacian constrained graphical model. In Advances in Neural Information Processing Systems (NeurIPS) , volume 34, pages 7101–7113, 2020. 26

    work page 2020

  48. [48]

    Online graph learning under smoothness priors

    Seyed Saman Saboksayr, Gonzalo Mateos, and M ¨ujdat C ¸ etin. Online graph learning under smoothness priors. In 2021 29th European Signal Processing Conference (EUSIPCO), pages 1820–1824. IEEE, 2021

    work page 2021

  49. [49]

    Graph signal recovery via primal-dual algorithms for total vari- ation minimization

    Pierre Berger, Gabor Hannak, and Gerald Matz. Graph signal recovery via primal-dual algorithms for total vari- ation minimization. IEEE Journal of Selected Topics in Signal Processing, 11(6):842–855, 2017

    work page 2017

  50. [50]

    Regular- ized Tyler’s scatter estimator: Existence, uniqueness, and algorithms

    Ying Sun, Prabhu Babu, and Daniel P Palomar. Regular- ized Tyler’s scatter estimator: Existence, uniqueness, and algorithms. IEEE Transactions on Signal Processing , 62(19):5143–5156, 2014

    work page 2014

  51. [51]

    Nonconvex sparse graph learning under Laplacian constrained graphical model

    Jiaxi Ying, Jos ´e Vin ´ıcius de Miranda Cardoso, and Daniel Palomar. Nonconvex sparse graph learning under Laplacian constrained graphical model. Advances in Neural Information Processing Systems , 33:7101–7113, 2020

    work page 2020

  52. [52]

    Network topol- ogy inference from non-stationary graph signals

    Rasoul Shafipour and Gonzalo Mateos. Network topol- ogy inference from non-stationary graph signals. In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pages 5564–5568, 2021

    work page 2021

  53. [53]

    Structural learn- ing with time-varying components: tracking the cross- section of financial time series

    Makram Talih and Nicolas Hengartner. Structural learn- ing with time-varying components: tracking the cross- section of financial time series. Journal of the Royal Statistical Society, Series B , 67(3):321–341, 2005

    work page 2005

  54. [54]

    Regular- ized estimation of piecewise constant Gaussian graphical models: the group-fused graphical Lasso

    Alexander J Gibberd and James D B Nelson. Regular- ized estimation of piecewise constant Gaussian graphical models: the group-fused graphical Lasso. Journal of Computational and Graphical Statistics , 26(3):623–634, 2017

    work page 2017

  55. [55]

    Factor modeling for high- dimensional time series: Inference for the number of factors

    Clifford Lam and Qiwei Yao. Factor modeling for high- dimensional time series: Inference for the number of factors. Annals of Statistics , 40(2):694–726, 2012

    work page 2012

  56. [56]

    Forecasting using principal components from a large number of predic- tors

    James H Stock and Mark W Watson. Forecasting using principal components from a large number of predic- tors. Journal of the American Statistical Association , 97(460):1167–1179, 2002

    work page 2002

  57. [57]

    Determining the number of factors in approximate factor models

    Jushan Bai and Serena Ng. Determining the number of factors in approximate factor models. Econometrica, 70(1):191–221, 2002

    work page 2002

  58. [58]

    Prob- abilistic principal component analysis

    Michael E Tipping and Christopher M Bishop. Prob- abilistic principal component analysis. Journal of the Royal Statistical Society, Series B, 61(3):611–622, 1999

    work page 1999

  59. [59]

    Optimization Algorithms on Matrix Mani- folds

    Pierre-Antoine Absil, Robert Mahony, and Rodolphe Sepulchre. Optimization Algorithms on Matrix Mani- folds. Princeton University Press, Princeton, NJ, 2008

    work page 2008

  60. [60]

    Stochastic gradient descent on Rie- mannian manifolds

    Silvere Bonnabel. Stochastic gradient descent on Rie- mannian manifolds. IEEE Transactions on Automatic Control, 58(9):2217–2229, 2013

    work page 2013

  61. [61]

    Conic geometric opti- mization on the manifold of positive definite matrices

    Suvrit Sra and Reshad Hosseini. Conic geometric opti- mization on the manifold of positive definite matrices. SIAM Journal on Optimization , 25(1):713–739, 2015

    work page 2015

  62. [62]

    Manopt, a Matlab toolbox for optimization on manifolds

    Nicolas Boumal, Bamdev Mishra, Pierre-Antoine Absil, and Rodolphe Sepulchre. Manopt, a Matlab toolbox for optimization on manifolds. Journal of Machine Learning Research, 15:1455–1459, 2014

    work page 2014

  63. [63]

    An Introduction to Optimization on Smooth Manifolds

    Nicolas Boumal. An Introduction to Optimization on Smooth Manifolds. Cambridge University Press, 2023

    work page 2023

  64. [64]

    Learning time-varying graphs from online data

    Alberto Natali, Elvin Isufi, Mario Coutino, and Geert Leus. Learning time-varying graphs from online data. IEEE Open Journal of Signal Processing , 3:212–228, 2022

    work page 2022

  65. [65]

    Online topology inference from streaming stationary graph signals with partial connectivity information

    Rasoul Shafipour and Gonzalo Mateos. Online topology inference from streaming stationary graph signals with partial connectivity information. Algorithms, 13(9):228, 2020

    work page 2020

  66. [66]

    Joint structural break detection and parameter estimation in high-dimensional nonstationary V AR models

    Abolfazl Safikhani and Ali Shojaie. Joint structural break detection and parameter estimation in high-dimensional nonstationary V AR models. Journal of the American Statistical Association, 117(537):251–264, 2022

    work page 2022

  67. [67]

    Optimal nonparametric multivariate change point detection and localization

    Oscar Hernan Madrid Padilla, Yi Yu, Daren Wang, and Alessandro Rinaldo. Optimal nonparametric multivariate change point detection and localization. IEEE Transac- tions on Information Theory , 68(3):1922–1944, 2021

    work page 1922

  68. [68]

    Regularized estima- tion and testing for high-dimensional multi-block vector- autoregressive models

    Jiahe Lin and George Michailidis. Regularized estima- tion and testing for high-dimensional multi-block vector- autoregressive models. Journal of Machine Learning Research, 18(117):1–45, 2017

    work page 2017

  69. [69]

    SDFV AE: Static and dynamic factorized V AE for anomaly detection of multivariate CDN metrics

    Defu Cao, Yujing Wang, Juanyong Duan, Ce Zhang, Xia Zhu, Congrui Huang, Yunhai Tong, Bixiong Xu, Jing Bai, Jie Tong, and Qi Zhang. SDFV AE: Static and dynamic factorized V AE for anomaly detection of multivariate CDN metrics. In The Web Conference (WWW), pages 3076–3086, 2021

    work page 2021

  70. [70]

    Graph neural network- based anomaly detection in multivariate time series

    Ailin Deng and Bryan Hooi. Graph neural network- based anomaly detection in multivariate time series. In Proceedings of the AAAI conference on artificial intelligence (AAAI), volume 35, pages 4027–4035, 2021

    work page 2021

  71. [71]

    Gated graph sequence neural networks

    Yaguang Li, Daniel Tarlow, Marc Brockschmidt, and Richard Zemel. Gated graph sequence neural networks. In International Conference on Learning Representa- tions (ICLR), 2016

    work page 2016

  72. [72]

    Dynamic Factor Models, Cointegration, and Error Correction Mechanisms

    Matteo Barigozzi, Marco Lippi, and Matteo Luciani. Dy- namic factor models, cointegration, and error correction mechanisms. arXiv preprint arXiv:1510.02399 , 2015

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  73. [73]

    On the estimation of integrated covariance matrices of high dimensional diffusion processes

    Xinghua Zheng and Yingying Li. On the estimation of integrated covariance matrices of high dimensional diffusion processes. The Annals of Statistics , pages 3121–3151, 2011

    work page 2011

  74. [74]

    A survey and comparison of contemporary algorithms for computing the matrix geometric mean

    Ben Jeuris, Raf Vandebril, and Bart Vandereycken. A survey and comparison of contemporary algorithms for computing the matrix geometric mean. Electronic Trans- actions on Numerical Analysis , 39:379–402, 2012

    work page 2012

  75. [75]

    Riemannian proximal gradient methods

    Wen Huang, Pierre-Antoine Absil, Kyle A Gallivan, and Paul Hand. Riemannian proximal gradient methods. Mathematical Programming, 194(1):371–413, 2022

    work page 2022

  76. [76]

    Zhang and Q

    X. Zhang and Q. Wang. Graph learning with low- rank and diagonal structures: A Riemannian geometric approach. In Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2025

    work page 2025

  77. [77]

    N. K. Chandra, P. Mueller, and A. Sarkar. Bayesian scal- able precision factor analysis for massive sparse Gaus- sian graphical models. arXiv preprint arXiv:2107.11316, 27 2021

    work page arXiv 2021

  78. [78]

    Hippert-Ferrer, F

    A. Hippert-Ferrer, F. Bouchard, A. Mian, T. Vayer, and A. Breloy. Learning graphical factor models with rie- mannian optimization. In Proceedings of ECML PKDD , pages 349–366. Springer, 2023

    work page 2023

  79. [79]

    Riemannian optimization using three different metrics for Hermitian PSD fixed-rank con- straints

    Shixin Zheng, Wen Huang, Bart Vandereycken, and Xiangxiong Zhang. Riemannian optimization using three different metrics for Hermitian PSD fixed-rank con- straints. Computational Optimization and Applications , 91(3):1135–1184, 2025

    work page 2025

  80. [80]

    Positive Definite Matrices

    Rajendra Bhatia. Positive Definite Matrices . Princeton University Press, 2009

    work page 2009

Showing first 80 references.