pith. sign in

arxiv: 2506.09776 · v3 · submitted 2025-06-11 · 🧮 math.OC · cs.SY· eess.SY

A Saddle Point Algorithm for Robust Data-Driven Factor Model Problems

Shabnam Khodakaramzadeh , Soroosh Shafiee , Gabriel de Albuquerque Gleizer , Peyman Mohajerin Esfahani This is my paper

Pith reviewed 2026-05-19 09:38 UTC · model grok-4.3

classification 🧮 math.OC cs.SYeess.SY
keywords factor modelsaddle-point optimizationlinear minimization oraclerobust data-driven optimizationfirst-order algorithmFrobenius normKullback-Leibler divergenceWasserstein distance
0
0 comments X

The pith

A first-order algorithm solves robust data-driven factor models by calling linear minimization oracles for three common uncertainty sets.

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

The paper formulates the factor model problem of recovering low-dimensional structure from high-dimensional data as a robust saddle-point optimization. It develops a first-order method that iterates by calling a linear minimization oracle on the dual function. Semi-closed-form expressions are supplied for the oracles under Frobenius-norm, Kullback-Leibler, and Gelbrich (Wasserstein) uncertainty, together with explicit Lipschitz constants that control the convergence rate. High-dimensional numerical tests show the procedure outperforming generic solvers.

Core claim

The robust data-driven factor model admits a saddle-point reformulation whose dual can be minimized by a first-order algorithm driven by a linear minimization oracle; semi-closed solutions (up to a scalar) exist for the Frobenius, KL, and Gelbrich oracles, and their Lipschitz constants are derived explicitly to guarantee convergence.

What carries the argument

Linear minimization oracle (LMO) on the dual function of the saddle-point formulation, which supplies the update direction for each first-order step.

If this is right

  • The method yields explicit convergence rates once the Lipschitz constants of the three chosen oracles are inserted.
  • Each oracle can be evaluated in closed form up to a one-dimensional scalar search, making per-iteration cost scale linearly with dimension.
  • The same framework applies to any uncertainty set whose LMO admits a semi-closed solution.
  • Numerical evidence indicates the approach scales to regimes where off-the-shelf solvers become impractical.

Where Pith is reading between the lines

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

  • The same LMO technique could be reused for other robust estimation tasks whose duals satisfy comparable regularity.
  • Replacing the three uncertainty sets with new ones would require only deriving their respective LMOs and Lipschitz constants.
  • The observed high-dimensional speed-up suggests the method may remain practical when the number of factors or observations grows further.
  • If the factor model is embedded inside a larger learning pipeline, the saddle-point iterates could serve as a differentiable layer.

Load-bearing premise

The dual function of the saddle-point reformulation must be Lipschitz continuous for the first-order algorithm to converge at the stated rate.

What would settle it

Running the algorithm on a high-dimensional factor-model instance whose dual is known to be non-Lipschitz or whose optimal value differs from a high-accuracy interior-point solution by more than the predicted error bound.

Figures

Figures reproduced from arXiv: 2506.09776 by Gabriel de Albuquerque Gleizer, Peyman Mohajerin Esfahani, Shabnam Khodakaramzadeh, Soroosh Shafiee.

Figure 1
Figure 1. Figure 1: Convergence error (32) of algorithm (11) with the projection Algorithm 1 Ground-truth covariance matrix generation: A pseudorandom tall matrix ΦTrue ∈ R n×r , and a pseudorandom diagonal PSD matrix DTrue ∈ R n×n which are the true factor loading matrix and the true covariance matrix of idiosyncratic noise are created using rand function in MATLAB with rng(1,’twister’) and rng(0), respectively. The r-dimens… view at source ↗
Figure 2
Figure 2. Figure 2: Estimation error (34) of the ground-truth ΣTrue the cases where d(Σ, Σ) is F(Σ b , Σ), KL(Σ b ||Σ), and G(Σ b , Σ), respectively. The performance observed b here validates the theoretical convergence result of Proposition 2.3 for all the distance functions. 4.3. Estimation of the ground-truth ΣTrue In this numerical study, the effect of the hyperparameter ε on the estimation error of the true covariance ma… view at source ↗
Figure 3
Figure 3. Figure 3: Computational time comparison of algorithm (11) and MOSEK d(Σ, Σ) corresponding to KL(Σ b ||Σ), although a distinguishable sweet spot has not been obser b ved, in 43% of the experiments, a slight improvement in the estimation of ΣTrue, compared to Σ, is observed. b 4.4. Execution time In this subsection, the execution time of the proposed first-order algorithm is investigated and compared with the off-the-… view at source ↗
read the original abstract

We study the factor model problem, which aims to uncover low-dimensional structures in high-dimensional datasets. Adopting a robust data-driven approach, we formulate the problem as a saddle-point optimization. Our primary contribution is a first-order algorithm that solves this reformulation by leveraging a linear minimization oracle (LMO). We further develop semi-closed form solutions (up to a scalar) for three specific LMOs, corresponding to the Frobenius norm, Kullback-Leibler divergence, and Gelbrich (aka Wasserstein) distance. The analysis includes explicit quantification of these LMOs' regularity conditions, notably the Lipschitz constants of the dual function, which govern the algorithm's convergence performance. Numerical experiments confirm our method's effectiveness in high-dimensional settings, outperforming standard off-the-shelf optimization solvers.

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

1 major / 2 minor

Summary. The manuscript formulates the factor model problem as a robust data-driven saddle-point optimization and develops a first-order algorithm that solves the reformulation via a linear minimization oracle (LMO). Semi-closed-form solutions (up to a scalar) are derived for three LMOs corresponding to the Frobenius norm, Kullback-Leibler divergence, and Gelbrich (Wasserstein) distance. The analysis provides explicit quantification of the LMOs' regularity conditions, including Lipschitz constants of the dual functions that control convergence rates. Numerical experiments are reported to demonstrate effectiveness and superiority over off-the-shelf solvers in high-dimensional settings.

Significance. If the claimed Lipschitz constants remain bounded independently of dimension, the work supplies a practical first-order method with explicit oracles and convergence guarantees for robust factor models, which could be useful in high-dimensional statistics and optimization. The semi-closed-form LMO solutions and the regularity analysis constitute concrete technical contributions; the high-dimensional numerical validation further supports applicability.

major comments (1)
  1. [Convergence analysis after Gelbrich LMO derivation] In the convergence analysis following the three LMO derivations (particularly the Gelbrich/Wasserstein case): the dual-function Lipschitz constant is asserted to be finite and is expressed in terms of covariance eigenvalues and the radius parameter. The derivation does not show that this constant remains O(1) rather than scaling with ambient dimension d. Because the algorithm's step-size choice and iteration complexity are governed by this constant, any d-dependent growth would invalidate the claimed convergence performance precisely in the high-dimensional regimes highlighted in the abstract and experiments.
minor comments (2)
  1. The abstract states that the LMOs admit 'semi-closed form solutions (up to a scalar)'; the manuscript should explicitly identify the scalar in each of the three cases and describe the one-dimensional root-finding procedure used to recover it.
  2. Notation for the factor model dimensions (e.g., number of factors versus ambient dimension) should be introduced once in §2 and used consistently in the LMO derivations and Lipschitz bounds.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the careful reading and the insightful comment on the convergence analysis. We respond to the major comment below.

read point-by-point responses

  1. Referee: In the convergence analysis following the three LMO derivations (particularly the Gelbrich/Wasserstein case): the dual-function Lipschitz constant is asserted to be finite and is expressed in terms of covariance eigenvalues and the radius parameter. The derivation does not show that this constant remains O(1) rather than scaling with ambient dimension d. Because the algorithm's step-size choice and iteration complexity are governed by this constant, any d-dependent growth would invalidate the claimed convergence performance precisely in the high-dimensional regimes highlighted in the abstract and experiments.

    Authors: We thank the referee for this observation. The Lipschitz constant for the Gelbrich dual function is derived explicitly as a function of the largest eigenvalue of the sample covariance and the uncertainty radius. In the factor-model setting the covariance admits a low-rank-plus-diagonal decomposition; under the standard assumptions used throughout the paper (bounded factor loadings and idiosyncratic variances), the operator norm of the covariance remains bounded independently of ambient dimension d. We will revise the manuscript to add an explicit remark (or short proposition) immediately after the Gelbrich LMO derivation that states and verifies this dimension-free bound on the Lipschitz constant. With this addition the step-size selection and iteration complexity remain independent of d, preserving the claimed high-dimensional performance. We view the revision as a clarification that strengthens the existing analysis rather than a change to the main results. revision: yes

Circularity Check

0 steps flagged

No circularity: algorithmic derivation and regularity analysis are independent of inputs

full rationale

The paper develops a first-order saddle-point algorithm using linear minimization oracles with explicitly derived semi-closed-form solutions for the Frobenius, KL, and Gelbrich cases, followed by direct quantification of dual-function Lipschitz constants from the problem data and radius parameters. These steps rely on standard convex analysis and oracle constructions applied to the given robust factor-model saddle-point reformulation; no step reduces a claimed result to a fitted parameter, self-citation chain, or definitional renaming. The convergence claims rest on the stated regularity conditions rather than on any self-referential construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Review based on abstract only; no explicit free parameters, ad-hoc axioms, or invented entities are stated.

axioms (1)
  • standard math Standard convex optimization assumptions hold for the saddle-point problem and the chosen divergence measures.
    The algorithm and LMO regularity analysis rely on convexity and Lipschitz continuity of the dual function.

pith-pipeline@v0.9.0 · 5682 in / 1175 out tokens · 48146 ms · 2026-05-19T09:38:41.910212+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

63 extracted references · 63 canonical work pages · 1 internal anchor

  1. [1]

    Ident ification of dynamic systems from noisy data: Single factor case

    Brian David Outram Anderson and Manfred Deistler. Ident ification of dynamic systems from noisy data: Single factor case. Mathematics of Control, Signals, and Systems , 6:10–29, 1993

    work page 1993

  2. [2]

    An Introduction to Multivariate Statistical Analysis

    Theodore Wilbur Anderson. An Introduction to Multivariate Statistical Analysis . Wiley, 2003

    work page 2003

  3. [3]

    Adaptivity of averaged stochastic gradie nt descent to local strong convexity for logistic regression

    Francis Bach. Adaptivity of averaged stochastic gradie nt descent to local strong convexity for logistic regression. Journal of Machine Learning Research , 15(1):595–627, 2014

    work page 2014

  4. [4]

    Determining the number of facto rs in approximate factor models

    Jushan Bai and Serena Ng. Determining the number of facto rs in approximate factor models. Econometrica, 70:191–221, 2002

    work page 2002

  5. [5]

    Convex Analysis and Monotone Operator Theory in Hilbert Spaces

    Heinz H Bauschke and Patrick L Combettes. Convex Analysis and Monotone Operator Theory in Hilbert Spaces. Springer, 2017

    work page 2017

  6. [6]

    Control of uncertain systems with a set-membership descript ion of the uncer- tainty

    Dimitri Bertsekas. Control of uncertain systems with a set-membership descript ion of the uncer- tainty. PhD thesis, Massachusetts Institute of Technology, 1971

    work page 1971

  7. [7]

    Convex Optimization Theory

    Dimitri Bertsekas. Convex Optimization Theory . Athena Scientific, 2009

    work page 2009

  8. [8]

    Convex Optimization Algorithms

    Dimitri Bertsekas. Convex Optimization Algorithms . Athena Scientific, 2015

    work page 2015

  9. [9]

    Copenhaver, and Rahul Maz umder

    Dimitris Bertsimas, Martin S. Copenhaver, and Rahul Maz umder. Certifiably optimal low rank factor analysis. Journal of Machine Learning Research , 18:1–53, 2017

    work page 2017

  10. [10]

    Strong con vexity of sandwiched entropies and related optimization problems

    Rajendra Bhatia, Tanvi Jain, and Yongdo Lim. Strong con vexity of sandwiched entropies and related optimization problems. Reviews in Mathematical Physics , 30(09):1850014, 2018

    work page 2018

  11. [11]

    Christopher M. Bishop. Pattern Recognition and Machine Learning . Springer, 2006

    work page 2006

  12. [12]

    The /suppress Lojasiewicz inequality for nonsmooth subanalytic functions with applications to subgradient dy namical systems

    J´ erˆ ome Bolte, Aris Daniilidis, and Adrian Lewis. The /suppress Lojasiewicz inequality for nonsmooth subanalytic functions with applications to subgradient dy namical systems. SIAM Journal on Optimization, 17(4):1205–1223, 2007

    work page 2007

  13. [13]

    Proximal alternating linearized minimization for nonconvex and nonsmooth problems

    J´ erˆ ome Bolte, Shoham Sabach, and Marc Teboulle. Proximal alternating linearized minimization for nonconvex and nonsmooth problems. Mathematical Programming, 146(1):459–494, 2014

    work page 2014

  14. [14]

    Springer, 2013

    Joseph Fr´ ed´ eric Bonnans and Alexander Shapiro.Perturbation Analysis of Optimization Problems . Springer, 2013

    work page 2013

  15. [15]

    Modeling complex sy stems by generalized factor analysis

    Giulio Bottegal and Giorgio Picci. Modeling complex sy stems by generalized factor analysis. IEEE Transactions on Automatic Control , 60(3):759–774, 2014

    work page 2014

  16. [16]

    Decisio n-theoretic planning: Structural assump- tions and computational leverage

    Craig Boutilier, Thomas Dean, and Steve Hanks. Decisio n-theoretic planning: Structural assump- tions and computational leverage. Journal of Artificial Intelligence Research , 11:1–94, 1999. A SADDLE POINT ALGORITHM FOR ROBUST DATA-DRIVEN F ACTOR MODE L PROBLEMS 20

    work page 1999

  17. [17]

    A method for finding p rojections onto the intersection of convex sets in Hilbert spaces

    James P Boyle and Richard L Dykstra. A method for finding p rojections onto the intersection of convex sets in Hilbert spaces. In Advances in Order Restricted Statistical Inference , pages 28–47. Springer, 1986

    work page 1986

  18. [18]

    Some matrix rearrangeme nt inequalities

    Eric Carlen and Elliott H Lieb. Some matrix rearrangeme nt inequalities. Annali di Matematica Pura ed Applicata , 185(Suppl 5):S315–S324, 2006

    work page 2006

  19. [19]

    An alternating minimization algorithm for Factor Analysis

    Valentina Ciccone, Augusto Ferrante, and Mattia Zorzi . An alternating minimization algorithm for factor analysis. arXiv:1806.04433, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  20. [20]

    Factor models with real data: A robust estimation of the number of factors

    Valentina Ciccone, Augusto Ferrante, and Mattia Zorzi . Factor models with real data: A robust estimation of the number of factors. IEEE Transactions on Automatic Control , 64(6):2412–2425, 2018

    work page 2018

  21. [21]

    Elements of Information Theory

    Thomas M Cover. Elements of Information Theory . Wiley, 1999

    work page 1999

  22. [22]

    A robust op- timization approach to network control using local informa tion exchange

    Georgios Darivianakis, Angelos Georghiou, Soroosh Sh afiee, and John Lygeros. A robust op- timization approach to network control using local informa tion exchange. Operations Research (Forthcoming), 2024

    work page 2024

  23. [23]

    The structure of generalized linear dynamic factor models

    Manfred Deistler, Wolfgang Scherrer, and Brian David O utram Anderson. The structure of generalized linear dynamic factor models. Empirical Economic and Financial Research: Theory, Methods and Practice , pages 379–400, 2015

    work page 2015

  24. [24]

    Delgado, Juan C

    Ram´ on A. Delgado, Juan C. Ag¨ uero, and Graham C. Goodwin. A rank-constrained optimization approach: Application to factor analysis. IF AC Proceedings Volumes, 47(3):10373–10378, 2014

    work page 2014

  25. [25]

    Maxim um likelihood from incomplete data via the EM algorithm

    Arthur P Dempster, Nan M Laird, and Donald B Rubin. Maxim um likelihood from incomplete data via the EM algorithm. Journal of the Royal Statistical Society: B , 39(1):1–22, 1977

    work page 1977

  26. [26]

    The rate of convergence o f Dykstra’s cyclic projections algorithm: The polyhedral case

    Frank Deutsch and Hein Hundal. The rate of convergence o f Dykstra’s cyclic projections algorithm: The polyhedral case. Numerical Functional Analysis and Optimization , 15(5-6):537–565, 1994

    work page 1994

  27. [27]

    Compressed sensing

    David L Donoho. Compressed sensing. IEEE Transactions on Information Theory , 52(4):1289– 1306, 2006

    work page 2006

  28. [28]

    An algorithm for restricted least sq uares regression

    Richard L Dykstra. An algorithm for restricted least sq uares regression. Journal of the American Statistical Association, 78(384):837–842, 1983

    work page 1983

  29. [29]

    A one-factor multivariat e time series model of metropolitan wage rates

    Robert Engle and Mark Watson. A one-factor multivariat e time series model of metropolitan wage rates. Journal of the American Statistical Association , 76(376):774–781, 1981

    work page 1981

  30. [30]

    A ro bust approach to arma factor modeling

    Lucia Falconi, Augusto Ferrante, and Mattia Zorzi. A ro bust approach to arma factor modeling. IEEE Transactions on Automatic Control , 69(2):828–841, 2023

    work page 2023

  31. [31]

    Large co variance estimation by thresholding principal orthogonal complements

    Jianqing Fan, Yuan Liao, and Martina Mincheva. Large co variance estimation by thresholding principal orthogonal complements. Journal of the Royal Statistical Society: B , 75(4):603–680, 2013

    work page 2013

  32. [32]

    The generalized dynamic-factor model: Identification and estimation

    Mario Forni, Marc Hallin, Marco Lippi, and Lucrezia Rei chlin. The generalized dynamic-factor model: Identification and estimation. Review of Economics and Statistics , 82(4):540–554, 2000

    work page 2000

  33. [33]

    The generalized dynamic fa ctor model: Representation theory

    Mario Forni and Marco Lippi. The generalized dynamic fa ctor model: Representation theory. Econometric Theory, 17(6):1113–1141, 2001

    work page 2001

  34. [34]

    A cyclic projection alg orithm via duality

    Norbert Gaffke and Rudolf Mathar. A cyclic projection alg orithm via duality. Metrika, 36(1):29– 54, 1989

    work page 1989

  35. [35]

    On a formula for the L2 wasserstein metric between measures on Euclidean and Hilbert spaces

    Matthias Gelbrich. On a formula for the L2 wasserstein metric between measures on Euclidean and Hilbert spaces. Mathematische Nachrichten , 147(1):185–203, 1990. A SADDLE POINT ALGORITHM FOR ROBUST DATA-DRIVEN F ACTOR MODE L PROBLEMS 21

    work page 1990

  36. [36]

    A class of Wassers tein metrics for probability distribu- tions

    Clark R Givens and Rae Michael Shortt. A class of Wassers tein metrics for probability distribu- tions. Michigan Mathematical Journal , 31(2):231–240, 1984

    work page 1984

  37. [37]

    A successive projection method

    Shih-Ping Han. A successive projection method. Mathematical Programming, 40(1):1–14, 1988

    work page 1988

  38. [38]

    System identifi cation by dynamic factor models

    Christiaan Heij and Wolfgang Scherrer. System identifi cation by dynamic factor models. SIAM Journal on Control and Optimization , 35(6):1924–1951, 1997

    work page 1924

  39. [39]

    A para metric estimation method for dynamic factor models of large dimensions

    George Kapetanios and Massimiliano Marcellino. A para metric estimation method for dynamic factor models of large dimensions. Journal of Time Series Analysis , 30(2):208–238, 2009

    work page 2009

  40. [40]

    Computation of the m aximum likelihood estimator in low-rank factor analysis

    Koulik Khamaru and Rahul Mazumder. Computation of the m aximum likelihood estimator in low-rank factor analysis. Mathematical Programming, 176:279–310, 2019

    work page 2019

  41. [41]

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

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

    work page 2012

  42. [42]

    Algorithms for non-ne gative matrix factorization

    Daniel Lee and H Sebastian Seung. Algorithms for non-ne gative matrix factorization. In Advances in neural information processing systems , pages 535–541, 2000

    work page 2000

  43. [43]

    Yalmip: A toolbox for modeling and optim ization in matlab

    Johan Lofberg. Yalmip: A toolbox for modeling and optim ization in matlab. In International Conference on Robotics and Automation , pages 284–289, 2004

    work page 2004

  44. [44]

    Error bounds and convergen ce analysis of feasible descent methods: A general approach

    Zhi-Quan Luo and Paul Tseng. Error bounds and convergen ce analysis of feasible descent methods: A general approach. Annals of Operations Research , 46(1):157–178, 1993

    work page 1993

  45. [45]

    The EM Algorithm and Extensions

    Geoffrey J McLachlan and Thriyambakam Krishnan. The EM Algorithm and Extensions . Wiley, 2007

    work page 2007

  46. [46]

    Mosek modeling cookbook: Release 3.3.1, 202 5

    Mosek ApS. Mosek modeling cookbook: Release 3.3.1, 202 5

    work page

  47. [47]

    Approximate prima l solutions and rate analysis for dual subgradient methods

    Angelia Nedi´ c and Asuman Ozdaglar. Approximate prima l solutions and rate analysis for dual subgradient methods. SIAM Journal on Optimization , 19(4):1757–1780, 2009

    work page 2009

  48. [48]

    Subgradient metho ds for saddle-point problems

    Angelia Nedi´ c and Asuman Ozdaglar. Subgradient metho ds for saddle-point problems. Journal of Optimization Theory and Applications , 142:205–228, 2009

    work page 2009

  49. [49]

    Distributionally robust inverse covariance estimation: The Wasserstein shrinkage estimat or

    Viet Anh Nguyen, Daniel Kuhn, and Peyman Mohajerin Esfa hani. Distributionally robust inverse covariance estimation: The Wasserstein shrinkage estimat or. Operations Research, 70:490–515, 2021

    work page 2021

  50. [50]

    Bridging bayesian and minimax mean square error estimation via Wasserstein distributionally robust optimization

    Viet Anh Nguyen, Soroosh Shafieezadeh-Abadeh, Daniel K uhn, and Peyman Mohajerin Esfahani. Bridging bayesian and minimax mean square error estimation via Wasserstein distributionally robust optimization. Mathematics of Operations Research , 48(1):1–37, 2023

    work page 2023

  51. [51]

    Hidden factor estimation in dynamic generalized factor analysis models

    Giorgio Picci, Lucia Falconi, Augusto Ferrante, and Ma ttia Zorzi. Hidden factor estimation in dynamic generalized factor analysis models. Automatica, 149:110834, 2023

    work page 2023

  52. [52]

    Guar anteed minimum-rank solutions of linear matrix equations via nuclear norm minimization

    Benjamin Recht, Maryam Fazel, and Pablo A Parrilo. Guar anteed minimum-rank solutions of linear matrix equations via nuclear norm minimization. SIAM Review , 52(3):471–501, 2010

    work page 2010

  53. [53]

    Stochastic gradient descen t for non-smooth optimization: Conver- gence results and optimal averaging schemes

    Ohad Shamir and Tong Zhang. Stochastic gradient descen t for non-smooth optimization: Conver- gence results and optimal averaging schemes. In International Conference on Machine Learning , pages 71–79. PMLR, 2013

    work page 2013

  54. [54]

    General intelligence,

    Charles Spearman. “General intelligence,” objective ly determined and measured. The American Journal of Psychology , 15(2):201–292, 1904

    work page 1904

  55. [55]

    Dykstra’s algorithm, ADMM, and coor dinate descent: Connections, insights, and extensions

    Ryan J Tibshirani. Dykstra’s algorithm, ADMM, and coor dinate descent: Connections, insights, and extensions. In Advances in Neural Information Processing Systems , 2017. A SADDLE POINT ALGORITHM FOR ROBUST DATA-DRIVEN F ACTOR MODE L PROBLEMS 22

    work page 2017

  56. [56]

    A coordinate gradient desc ent method for nonsmooth separable minimization

    Paul Tseng and Sangwoon Yun. A coordinate gradient desc ent method for nonsmooth separable minimization. Mathematical Programming, 117:387–423, 2009

    work page 2009

  57. [57]

    Stochastic realization prob lems motivated by econometric modeling

    Jan Hendrik van Schuppen. Stochastic realization prob lems motivated by econometric modeling. In Modelling, Identification and Robust Control , pages 259–275. 1986

    work page 1986

  58. [58]

    On rings of operators

    John von Neumann. On rings of operators. Reduction theo ry. Annals of Mathematics , 50(2):401– 485, 1949

    work page 1949

  59. [59]

    Convergence rate analy sis of a Dykstra-type projection algorithm

    Xiaozhou Wang and Ting Kei Pong. Convergence rate analy sis of a Dykstra-type projection algorithm. SIAM Journal on Optimization , 34(1):563–589, 2024

    work page 2024

  60. [60]

    Barry M Wise, Neal B Gallagher, Stephanie Watts Butler, Daniel D White, and Gabriel G Barna. A comparison of principal component analysis, multiway pri ncipal component analysis, trilinear decomposition and parallel factor analysis for fault detec tion in a semiconductor etch process. Journal of Chemometrics , 13(3-4):379–396, 1999

    work page 1999

  61. [61]

    Factor-analysis based anom aly detection and clustering

    Ningning Wu and Jing Zhang. Factor-analysis based anom aly detection and clustering. Decision Support Systems , 42(1):375–389, 2006

    work page 2006

  62. [62]

    Mul tirate factor analysis models for fault detection in multirate processes

    Le Zhou, Yaoxin Wang, Zhiqiang Ge, and Zhihuan Song. Mul tirate factor analysis models for fault detection in multirate processes. IEEE Transactions on Industrial Informatics , 15(7):4076–4085, 2018

    work page 2018

  63. [63]

    Factor analysis o f moving average processes

    Mattia Zorzi and Rodolphe Sepulchre. Factor analysis o f moving average processes. In European Control Conference, pages 3579–3584, 2015

    work page 2015