Pith. sign in

REVIEW 4 major objections 6 minor 59 references

Splitting-based randomized dynamical low-rank approximations for stiff matrix differential equations

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Randomized low-rank integrators reach order one and two on stiff matrix ODEs

desk verdict A sensible combination of exponential splitting and randomized DLRA with honest experiments, but the B-step closure in Eq. (2.5) is unanalyzed and should gate acceptance. read the letter →

arxiv 2506.15259 v3 pith:A6W62TLP submitted 2025-06-18 math.NA cs.NA

Zi Wu , Yong-Liang Zhao , Xian-Ming Gu This is my paper · ORCID
classification math.NAcs.NA MSC 65F5568W2065L04
keywords low-ranksplittingdynamicalapproximationrank-adaptivematrixdifferentialequationRiccatirandomizednumericallinearalgebraLie-TrotterStrang
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper proposes and tests a family of low-rank integrators for stiff semilinear matrix differential equations of the form $X'(t) = AX(t) + X(t)A^\top + F(t,X(t))$. The core idea is to split the equation into a stiff linear subproblem, solved exactly with matrix exponentials, and a nonstiff nonlinear subproblem, advanced by randomized dynamical low-rank methods that sketch the evolving row and column spaces. The paper claims that the Lie-Trotter composition gives first-order accuracy, the Strang composition gives second-order accuracy, and that both extend to rank-adaptive schemes, with numerical evidence on Allen-Cahn and differential Riccati problems. A sympathetic reader would care because existing projector-splitting low-rank integrators for stiff problems are sensitive to small singular values and expensive, while this approach is designed to remove step-size restrictions and avoid forming full-rank matrices.

What carries the argument

The central object is the pair of randomized dynamical low-rank solvers, DRSVD and DGN, built on a dynamical randomized range finder. The range finder advances the sketched ODE $\dot B(t) = F(t, B(t)(\Omega^\top\Omega)^{-1}\Omega^\top)\Omega$ (and its transpose analogue) to estimate the dominant column space of the nonlinear flow, then post-processes by solving reduced ODEs on the augmented bases $Q = \mathrm{orth}([U_0, Q_\tau])$; DRSVD projects the nonlinear ODE onto that basis, while DGN additionally samples the row space and couples three small ODEs. The splitting with the exactly integrated linear flow is what keeps the solution on the rank-$r$ manifold, and the randomized steps are what make the nonlinear subproblem cheap to advance.

What would settle it

Take a small semilinear matrix ODE with a known exact solution, run one step of the nonlinear subproblem, and compare the true $N(t_0+\tau)$ with the sketched reconstruction $B(t_0+\tau)(\Omega^\top\Omega)^{-1}\Omega^\top$ for a few Gaussian sketches of increasing oversampling $p$; if this reconstruction error fails to decay with $p$ or $\tau$, or if the overall method's error is dominated by this term, the range-inclusion premise fails and the claimed convergence orders do not hold in general.

Watch

Extended reading notes

Core claim

The paper's central claim is that operator splitting and randomized dynamical low-rank approximation can be combined into practical integrators for (1.1). The stiff linear flow $\Phi^A_\tau(M_0) = e^{\tau A}M_0 e^{\tau A^\top}$ preserves rank exactly, so the approximation stays on the rank-$r$ manifold; the nonlinear flow $\Phi^F_\tau$ is approximated by a dynamic randomized range finder followed by either a dynamic randomized SVD (DRSVD) or a dynamic generalized Nyström (DGN) post-processing step that solves only small sketched ODEs. Composing these as $\widetilde L_\tau = \Phi^A_\tau \circ \widetilde\Phi^F_\tau$ yields a first-order Lie-Trotter method and $\widetilde S_\tau = \Phi^A_{\tau/2} \circ \widetilde\Phi^F_\tau \circ \Phi^A_{\tau/2}$ yields a second-order Strang method, with rank-adaptive variants based on tolerance-driven truncation. On the Allen-Cahn equation discretized on $1024\times1024$ grids and on a differential Riccati equation, the methods show the expected temporal orders and lower relative errors than the low-rank projector-splitting reference.

Load-bearing premise

The whole convergence story depends on the randomized range finder tracking the true dominant row and column spaces of the nonlinear flow closely enough that replacing $F(t,N(t))$ with $F(t, B(t)(\Omega^\top\Omega)^{-1}\Omega^\top)$ (and the analogous post-processing projection) introduces errors smaller than the splitting error; the paper does not state or bound the error from the unsketched complement.

Editorial extensions

If this is right

  • The Lie-Trotter variant converges at order one and the Strang variant at order two in time on stiff semilinear matrix ODEs of the form (1.1).
  • Because the stiff linear part is integrated exactly by matrix exponentials, the methods avoid explicit CFL-type step-size restrictions and are suited to long-time simulations.
  • Both DRSVD and DGN avoid forming full-rank matrices in the nonlinear step, working with sketched matrices of size $m \times (r+p)$ and related small cores, so the cost scales with the chosen rank rather than with the full state dimension.
  • The rank-adaptive extensions replace fixed-rank truncation by tolerance-based truncation, so users do not need to preselect the target rank.

Reading between the lines

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

  • A natural extension the paper leaves implicit is to monitor the sketched residual $\|B(t)-QQ^\top B(t)\|_F$ during the nonlinear update and restart or enrich the sketch when it grows, which would guard against the unstated range-inclusion error behind Eq. (2.5).
  • Because the composition is symmetric, higher-order methods could likely be built by composing the same Strang step with weighted coefficients, although the paper does not attempt this.
  • The authors note in the concluding remarks that truncated SVD may become unstable for large initial perturbations; a concrete extension would be to swap the SVD truncation for a regularized or QLP-based truncation, or to combine the adaptive rank selection with a restart strategy.
  • The DGN variant's double sketching of both row and column spaces should make it preferable for nonsymmetric or rectangular problems, an advantage not directly tested by the symmetric Allen-Cahn and Riccati examples.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes splitting-based randomized dynamical low-rank (RDLR) approximations for stiff semilinear matrix differential equations of the form X'(t) = AX(t) + X(t)A^T + F(t,X(t)). The stiff linear part is integrated exactly with matrix exponentials, while the nonstiff nonlinear part is approximated by the randomized dynamical low-rank solvers DRSVD and DGN due to Carrel. Lie-Trotter and Strang compositions are formulated, and rank-adaptive variants are introduced. Numerical experiments on semi-discretized Allen-Cahn equations and differential Riccati equations report first-order (Lie-Trotter) and second-order (Strang) temporal convergence when the target rank is sufficiently large, together with spatial convergence studies and long-time interface simulations.

Significance. If the convergence claims are correct, the paper offers a potentially efficient alternative to projector-splitting low-rank integrators for stiff matrix-valued ODEs, combining operator splitting with randomized range finding. The manuscript is usefully explicit: Algorithms 1 and 2 and A.3-A.6 give complete pseudocode, and the tables report errors over many ranks, step sizes, and grid sizes, which allows a reader to see the rank dependence directly. However, the central claim is empirical: no error bounds are proved, a key range-closure assumption in the B-step is unstated, and the observed convergence order degrades at low ranks. The contribution is therefore best read as a numerical demonstration whose robustness conditions still need to be pinned down.

major comments (4)
  1. [§2.3.1, Eq. (2.5)] The B-step closes the nonlinearity by replacing N(t) with B(t)(Ω^T Ω)^{-1} Ω^T, which is exact only if Range(N(t)^T) is contained in Range(Ω) for all t in the substep; this condition is never stated and no bound is given for the projection error. The argument preceding Eq. (2.5) establishes Range(B(t0+τ)) = Range(N(t0+τ)) only at the endpoint and only under the rank assumption rank(N(t0+τ)) ≤ r+p, not for intermediate times. The same unverified condition enters Algorithm A.3 (lines 2 and 7), Algorithm A.4 (step 3), and the exactness claims imported from [8, Lemmas 3.1 and 3.2]. Because every convergence table in Section 4 is produced through this substep, the observed first- and second-order rates cannot be separated from the assumption that the random sketch tracks the row and column spaces of the flow; Section 5 explicitly defers an error analysis.
  2. [§4.1.1, Tables 4.1-4.2; §4.1.2, Tables 4.7-4.8] The observed convergence orders degrade or become negative at low ranks, so the claimed 'desired convergence orders' are not a uniform property of the methods. In Table 4.2, DRSVD-ST at r=12 gives rates -0.0942 and -0.1169 for M=64 and 128, respectively; in Table 4.8, DRSVD-ST at r=4 gives rates -0.0924 and -0.1083 for M=256 and 512. Even the Lie-Trotter variants saturate at low ranks, e.g., Table 4.1 with r=12 shows a rate of only 0.4351 between M=128 and 256, and Table 4.7 with r=4 shows 0.3670 between M=256 and 512. The paper should either quantify the rank threshold above which the nominal orders hold or explicitly qualify the convergence statements in the abstract and introduction.
  3. [§4.1, Algorithms A.3-A.5] The numerical experiments never report the oversampling parameters p (and ℓ for DGN) or the number of power iterations q used in the randomized range finders, although these are free inputs to every method. Without these values the convergence tables cannot be reproduced, and the possibility that the observed high-order rates depend on generous oversampling or power iterations cannot be assessed. Please report the parameter values used in Tables 4.1-4.9 and, ideally, include a sensitivity study with respect to p and q.
  4. [§5, item (ii)] The manuscript contains no error analysis for either the randomized range finder or the composed splitting scheme; Section 5 item (ii) acknowledges this gap. Because the nonlinear substep is approximated by a stochastic low-rank solver, the standard Lie-Trotter and Strang splitting order results do not automatically transfer to the composed method. At minimum, the authors should state the conditions (e.g., invariant subspaces, Lipschitz constants, and rank requirements) under which first- and second-order convergence is expected, and should clearly distinguish measured numerical rates from proven rates.
minor comments (6)
  1. [§2.3.1, after Eq. (2.5)] The equation is described as 'autonomous' although F depends explicitly on t; this wording should be corrected to avoid confusion.
  2. [Algorithm A.6, line 3] The formula F(t, B(t)Ω^T Ω)^{-1} Ω^T) Ω has mismatched parentheses; it should read F(t, B(t)(Ω^T Ω)^{-1} Ω^T) Ω.
  3. [Tables 4.1-4.8] Several numerical entries contain typographical artifacts such as '1.66 04E-5', '1.9 162E-6', and '1.6 604E-5'; these should be cleaned before publication.
  4. [Fig. 4.1 caption] The caption states both 'different time step size τ' and 'the number of time steps M = 1024'; these are inconsistent and the caption should clarify whether the abscissa is τ or M.
  5. [§4.1.2] The control system writes x ∈ R^m while A ∈ R^{d×d}; the notation for the state dimension should be made consistent.
  6. [Algorithm A.5, line 3] The initial condition B(t0) = A0 Q2 should presumably be B(t0) = N0 Q2, since A0 is not defined in the algorithm.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: convergence orders are measured against reference solutions, and the imported randomized low-rank machinery is external support, not a self-referential fit.

full rationale

After walking the claimed derivation chain, I find no circular step. The method splits (1.1) into a stiff linear subproblem (2.1) and a nonstiff nonlinear subproblem (2.2), solves the linear part exactly with matrix exponentials, and applies the randomized dynamical low-rank methods DRSVD/DGN from Carrel [8] to the nonlinear part. The exactness conditions for those methods are stated as range-inclusion assumptions: 'as established in [8, Lemma 3.1], when the system satisfies the regularity condition Range(N(t)) ⊂ Range([U0,Qτ]) = Range(Q), for all t ∈ [t0,t0+τ], Eq. (2.7) becomes exact reconstruction of the solution' (Section 2.3.2). These are external lemmas, not results proved in this paper, and they are not used to define the target quantities. The claimed first- and second-order convergence rates are empirical measurements reported in Tables 4.1-4.10 against reference solutions (dop54 for the Riccati case, relative Frobenius-norm errors), not outputs of fitted parameters. The main caveat is that the B-step closure (2.5), which replaces N(t) by B(t)(Ω^TΩ)^{-1}Ω^T, is exact only under a row-space inclusion condition that is not stated or bounded, and Section 5 explicitly lists 'establishment of a systematic error analysis framework for the proposed low-rank numerical scheme' as future work. This is a rigor/correctness limitation, not circularity: no equation in the paper is equivalent to its own input by construction, and the self-citations [57,58] are contextual rather than load-bearing. Therefore the circularity score is 0.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The method imports its randomized machinery wholesale from Carrel [8] and the splitting from the exponential integrator/splitting literature. No new physical or mathematical entity is introduced. The user-selected ranks, tolerances, and unstated oversampling/power-iteration values affect the reported convergence behavior, but they are not fitted to the target output.

free parameters (4)
  • target rank r = 12, 14, 16, 18 (Allen-Cahn); 4, 6, 8, 10 (DRE)
    Chosen per experiment; the observed convergence orders only materialize for sufficiently large r (e.g., r >= 16 for Strang on Allen-Cahn).
  • oversampling p = not reported
    Required by Algorithms A.3-A.5 but never specified for the numerical experiments, so the results are not fully reproducible.
  • power iterations q = not reported
    Required by the dynamical range finder; never specified.
  • adaptive tolerances = rel 1e-8, abs 1e-12, range finder 1e-8
    Hand-chosen tolerances for the rank-adaptive variants in Tables 4.3 and 4.9.
assumptions (5)
  • standard math Variation-of-constants formula represents the exact solution of (1.1).
    Invoked at the start of Section 2 to justify the splitting approach.
  • standard math The linear flow X'=AX+XA^T preserves the rank-r manifold.
    Used in Section 2.2 to argue the stiff substep needs no rank truncation.
  • domain assumption Lie-Trotter and Strang splitting preserve their classical first/second order when the sub-flows are approximated accurately.
    Standard splitting theory [21,35] is imported without re-derivation; the paper gives no combined error bound for the randomized sub-flows.
  • domain assumption The randomized dynamical low-rank methods from [8] are accurate under the range-inclusion conditions (Range(N(t)) subset of Range(Q1), corange subset of Range(Q2)).
    Relied on in Section 2.3.2-2.3.3; exactness lemmas are quoted from [8] and not proved in this paper.
  • domain assumption The nonlinear term F is nonstiff and the stiffness is entirely in the linear term AX+XA^T.
    The splitting trick in Section 2.1 depends on this separation for the claimed stability and step-size behavior.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Splitting-based randomized dynamical low-rank approximations for stiff matrix differential equations." pith.science (2026). https://pith.science/paper/A6W62TLP

@misc{pith2026250615259,
  author       = {Pith},
  title        = {Pith review of: Splitting-based randomized dynamical low-rank approximations for stiff matrix differential equations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/A6W62TLP}},
  note         = {Machine review of arXiv:2506.15259}
}
read the original abstract

In the fields of control theory and machine learning, the dynamic low-rank approximation for large-scale matrices has received substantial attention. Considering large-scale semilinear stiff matrix differential equations, we propose splitting-based randomized dynamical low-rank approximations for a low-rank solution of the stiff matrix differential equation. We first split such the equation into a stiff linear subproblem and a nonstiff nonlinear subproblem. Then, a low-rank exponential integrator is applied to the linear subproblem. Two randomized low-rank approaches are employed for the nonlinear subproblem. Furthermore, we extend the proposed methods to rank-adaptation scenarios. Through rigorous validation on canonical stiff matrix differential problems, including spatially discretized Allen-Cahn equations and differential Riccati equations, we demonstrate that our methods achieve desired convergence orders. Numerical results confirm the robustness and accuracy of the proposed methods.

Figures

Figures reproduced from arXiv: 2506.15259 by the authors.

Figure 4
Figure 4. demonstrates that Eq. (4.1) solved at low rank [PITH_FULL_IMAGE:figures/full_fig_p014_4.png] view at source ↗
Figure 4
Figure 4. [PITH_FULL_IMAGE:figures/full_fig_p015_4.png] view at source ↗
Figure 4
Figure 4. [PITH_FULL_IMAGE:figures/full_fig_p016_4.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: illustrates the temporal evolution of errors for bo [PITH_FULL_IMAGE:figures/full_fig_p017_4.png]
Figure 4
Figure 4. Figure 4 [PITH_FULL_IMAGE:figures/full_fig_p018_4.png]
Figure 4
Figure 4. Figure 4: illustrates the temporal evolution of errors for Li [PITH_FULL_IMAGE:figures/full_fig_p021_4.png]
Figure 4
Figure 4. Figure 4 [PITH_FULL_IMAGE:figures/full_fig_p022_4.png]
Figure 4
Figure 4. Figure 4 [PITH_FULL_IMAGE:figures/full_fig_p023_4.png]
Figure 4
Figure 4. Figure 4: presents numerical simulations of the dynamic evol [PITH_FULL_IMAGE:figures/full_fig_p024_4.png]
Figure 4
Figure 4. Figure 4 [PITH_FULL_IMAGE:figures/full_fig_p025_4.png]
Figure 4
Figure 4. Figure 4 [PITH_FULL_IMAGE:figures/full_fig_p026_4.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

59 extracted references · 53 canonical work pages

  1. [8]

    arXiv preprint arXiv:2410.17091, 24 pages (2024)

    Carrel, B.: Randomised methods for dynamical low-rank ap proximation. arXiv preprint arXiv:2410.17091, 24 pages (2024)

    arXiv 2024

  2. [1]

    Birkh¨ auser, Basel, Switzerland (2012)

    Abou-Kandil, H., Freiling, G., Ionescu, V ., Jank, G.: Mat rix Riccati Equations in Control and Systems Theory. Birkh¨ auser, Basel, Switzerland (2012)

    work page 2012

  3. [2]

    Al-Mohy, A.H., Higham, N.J.: Computing the action of the m atrix exponential, with an application to exponential integrators. SIAM J. Sci. Comput. 33, 488–511 (2011)

    work page 2011

  4. [3]

    Acta Metall

    Allen, S.M., Cahn, J.W.: Ground state structures in order ed binary alloys with second neighbor inter- actions. Acta Metall. 20, 423–433 (1972) Split-Rand DLRA for Stiff MDEs 27 Algorithm A.3 Dynamical Rangefinder DynamicalRangefinder(N0,τ,F,r, p,q) Require: Initial value N0 ∈ Rm×n, step size τ > 0, target rank r > 0, oversampling parameters p ≥ 0, power i...

    work page 1972

  5. [4]

    Allen, S.M., Cahn, J.W.: A correction to the ground state o f FCC binary ordered alloys with first and second neighbor pairwise interactions. Scr. Metall. 7, 1261–1264 (1973)

    work page 1973

  6. [5]

    Courier Corporation, North Chelmsford, MA (2007)

    Anderson, B.D., Moore, J.B.: Optimal Control: Linear Qua dratic Methods. Courier Corporation, North Chelmsford, MA (2007)

    work page 2007

  7. [6]

    IEEE Trans

    Benner, P ., Mena, H.: Rosenbrock methods for solving Ricc ati differential equations. IEEE Trans. Automat. Contr. 58, 2950–2956 (2018)

    work page 2018

  8. [7]

    Caliari, M., Kandolf, P ., Ostermann, A., Rainer, S.: The L eja method revisited: backward error analy- sis for the matrix exponential. SIAM J. Sci. Comput. 38, A1639–A1661 (2016)

    work page 2016

Show all 59 references
  1. [9]

    BIT 62, 1149–1174 (2022)

    Ceruti, G., Kusch, J., Lubich, C.: A rank-adaptive robust integrator for dynamical low-rank approxi- mation. BIT 62, 1149–1174 (2022)

    2022

  2. [10]

    In: Proceedings of the Forty-First Annual ACM Symposium on Theory of Computing , pp

    Clarkson, K.L., Woodruff, D.P .: Numerical linear algeb ra in the streaming model. In: Proceedings of the Forty-First Annual ACM Symposium on Theory of Computing , pp. 205–214. ACM, New Y ork, NY (2009) 28 Zi Wu et al. Algorithm A.5 Dynamical Generalised Nystr¨ om DynamicalGen...

    2009

  3. [11]

    Czech Math

    Collegari, R., Federson, M., Frasson, M.: Linear FDEs in the frame of generalized ODEs: variation- of-constants formula. Czech Math. J. 68, 889–920 (2018)

    2018

  4. [12]

    Springer In- ternational Publishing, Cham, Switzerland (2016)

    Darwish, F.M.L.M.M.: The Finite V olume Method in Comput ational Fluid Dynamics. Springer In- ternational Publishing, Cham, Switzerland (2016)

    2016

  5. [13]

    Dormand, J.R., Prince, P .J.: A family of embedded Runge- Kutta formulae. J. Comput. Appl. Math. 6, 19–26 (1980)

    1980

  6. [14]

    part 1: Dirichlet boundary conditions

    Einkemmer, L., Ostermann, A.: Overcoming order reducti on in diffusion-reaction splitting. part 1: Dirichlet boundary conditions. SIAM J. Sci. Comput. 37, A1577–A1592 (2015)

    2015

  7. [15]

    IEEE Trans

    Fowlkes, C., Belongie, S., Chung, F., Malik, J.: Spectra l grouping using the Nystr¨ om method. IEEE Trans. Pattern Anal. Mach. Intell. 26, 214–225 (2004)

    2004

  8. [16]

    Gittens, A., Mahoney, M.W.: Revisiting the Nystr¨ om method for improved large-scale machine learn- ing. J. Mach. Learn. Res. 17, 3977–4041 (2016)

    2016

  9. [17]

    GAMM-Mitt

    G¨ uttel, S.: Rational Krylov approximation of matrix fu nctions: Numerical methods and optimal pole selection. GAMM-Mitt. 36, 8–31 (2013)

    2013

  10. [18]

    Springer, Berlin, Heidelberg (1993)

    Hairer, E., Wanner, G., Nørsett, S.P .: Solving Ordinary Differential Equations I: Nonstiff problems. Springer, Berlin, Heidelberg (1993)

    1993

  11. [19]

    SIAM Rev

    Halko, N., Martinsson, P .G., Tropp, J.A.: Finding struc ture with randomness: Probabilistic algorithms for constructing approximate matrix decompositions. SIAM Rev. 53, 217–288 (2011)

    2011

  12. [20]

    Springer Science & Business Media, London, UK (2012)

    Helmke, U., Moore, J.B.: Optimization and Dynamical Sys tems. Springer Science & Business Media, London, UK (2012)

    2012

  13. [21]

    Acta Numer

    Hochbruck, M., Ostermann, A.: Exponential integrators . Acta Numer. 19, 209–286 (2010) Split-Rand DLRA for Stiff MDEs 29 Algorithm A.6 Adaptive Dynamical Rangefinder AdaptiveDynamicalRangefinder(N0,τ,F,ϕ,β ) Require: Initial value N0 ∈ Rm×n, step size τ > 0, tolerance ϕ > 0, ma...

    2010

  14. [22]

    Maple Trans

    Jang, T., Uppaluri, M., Seo, K., Ramadesigan, V ., Subram anian, V .: A sparse differential algebraic equation (DAE) and stiff ordinary differential equation (O DE) solver in maple. Maple Trans. 3, No. 16,701 (2023). doi:10.5206/mt.v3i4.16701

    2023 doi

  15. [23]

    BMC Bioinformatics 17, 1–13 (2016)

    Kapur, A., Marwah, K., Alterovitz, G.: Gene expression p rediction using low-rank matrix completion. BMC Bioinformatics 17, 1–13 (2016)

    2016

  16. [24]

    Kassam, A.K., Trefethen, L.N.: Fourth-order time-step ping for stiff PDEs. SIAM J. Sci. Comput. 26, 1214–1233 (2005)

    2005

  17. [25]

    Kieri, E., Lubich, C., Walach, H.: Discretized dynamica l low-rank approximation in the presence of small singular values. SIAM J. Numer. Anal. 54, 1020–1038 (2016)

    2016

  18. [26]

    Kieri, E., V andereycken, B.: Projection methods for dyn amical low-rank approximation of high- dimensional problems. Comput. Meth. Appl. Math. 19, 73–92 (2019)

    2019

  19. [27]

    Kieri, E., V andereycken, B.: Stability and efficiency of dynamical low-rank integrators. SIAM J. Numer. Anal. 60, 1283–1308 (2022)

    2022

  20. [28]

    Koch, O., Lubich, C.: Dynamical low-rank approximation . SIAM J. Mat. Anal. Appl. 29, 434–454 (2007)

    2007

  21. [29]

    Lam, H.Y ., Ceruti, G., Kressner, D.: Randomized low-ran k Runge–Kutta methods. SIAM J. Mat. Anal. Appl. 46, 1587–1615 (2025)

    2025

  22. [30]

    Oxford University Press, Oxford, UK (1995)

    Lancaster, P ., Rodman, L.: Algebraic Riccati Equations . Oxford University Press, Oxford, UK (1995)

    1995

  23. [31]

    SIAM, Philadelphia, PA (2007)

    LeV eque, R.J.: Finite Difference Methods for Ordinary a nd Partial Differential Equations. SIAM, Philadelphia, PA (2007)

    2007

  24. [32]

    BIT 54, 171–188 (2014)

    Lubich, C., Oseledets, I.V .: A projector-splitting int egrator for dynamical low-rank approximation. BIT 54, 171–188 (2014)

    2014

  25. [33]

    arXiv preprint arXiv:1607.01649, 36 pages (2016)

    Martinsson, P .G.: Randomized methods for matrix comput ations. arXiv preprint arXiv:1607.01649, 36 pages (2016)

    2016 arXiv

  26. [34]

    Acta Numer

    Martinsson, P .G., Tropp, J.A.: Randomized numerical li near algebra: Foundations and algorithms. Acta Numer. 29, 403–572 (2020)

    2020

  27. [35]

    A cta Numer

    McLachlan, R.I., Quispel, G.R.W.: Splitting methods. A cta Numer. 11, 341–434 (2002)

    2002

  28. [36]

    Mu, L., Wang, J., Y e, X., Zhao, S.: A new weak Galerkin finit e element method for elliptic interface problems. J. Comput. Phys. 325, 157–173 (2016)

    2016

  29. [37]

    arXiv preprint arXiv:2009.11392, 33 pages (2020) 30 Zi Wu et al

    Nakatsukasa, Y .: Fast and stable randomized low-rank ma trix approximation. arXiv preprint arXiv:2009.11392, 33 pages (2020) 30 Zi Wu et al

    2020 arXiv

  30. [38]

    Acta Math

    Nystr¨ om, E.J.: ¨Uber die praktische auff¨ osung von integralgleichungen mit anwendungen auf randw- ertaufgaben. Acta Math. 54, 185–204 (1930)

    1930

  31. [39]

    Ostermann, A., Piazzola, C., Walach, H.: Convergence of a low-rank Lie–Trotter splitting for stiff matrix differential equations. SIAM J. Numer. Anal. 57, 1947–1966 (2019)

    2019

  32. [40]

    Soft ware package, Technical University of Chemnitz (2000)

    Penzl, T.: A MA TLAB Toolbox for Large Lyapunov and Riccat i Equations, Model Reduction Prob- lems, and Linear–Quadratic Optimal Control Problems. Soft ware package, Technical University of Chemnitz (2000). URL https://www.tu-chemnitz.de/sfb393/lyapack. L Y APACK User’s Guide,...

    2000

  33. [41]

    Saad, Y .: Analysis of some Krylov subspace approximations to the matrix exponential operator. SIAM J. Numer. Anal. 29, 209–228 (1992)

    1992

  34. [42]

    Saha, R., Srivastava, V ., Pilanci, M.: Matrix compressi on via randomized low rank and low precision factorization. Adv. Neural. Inf. Process. Syst. 36, 18,828–18,872 (2023)

    2023

  35. [43]

    , Ramabhadran, B.: Low-rank matrix factor- ization for deep neural network training with high-dimensi onal output targets

    Sainath, T.N., Kingsbury, B., Sindhwani, V ., Arisoy, E. , Ramabhadran, B.: Low-rank matrix factor- ization for deep neural network training with high-dimensi onal output targets. In: 2013 IEEE Inter- national Conference on Acoustics, Speech and Signal Proces sing, pp. 6655–66...

    2013

  36. [44]

    MIT Press, Cambridge, MA (2002)

    Sch¨ olkopf, B., Smola, A.J.: Learning with Kernels: Sup port V ector Machines, Regularization, Opti- mization, and Beyond. MIT Press, Cambridge, MA (2002)

    2002

  37. [45]

    Sch¨ afer, F., Owhadi, H.: Efficient time integration of s tiff nonlinear PDEs via the dynamical adaptive low-rank approximation method. SIAM J. Sci. Comput. 43, A1–A29 (2021)

    2021

  38. [46]

    I EEE Trans

    Stoica, P ., Babu, P .: Low-rank covariance matrix estima tion for factor analysis in anisotropic noise: Application to array processing and portfolio selection. I EEE Trans. Signal. Process. 71, 1699–1711 (2023)

    2023

  39. [47]

    Strang, G.: On the construction and comparison of differ ence schemes. SIAM J. Numer. Anal. 5, 506–517 (1968)

    1968

  40. [48]

    Sutti, M., V andereycken, B.: Implicit low-rank Riemann ian schemes for the time integration of stiff partial differential equations. J. Sci. Comput. 101, No. 3 (2024). doi:10.1007/s10915-024-02629-8

    2024 doi

  41. [49]

    Tropp, J.A., Y urtsever, A., Udell, M., Cevher, V .: Pract ical sketching algorithms for low-rank matrix approximation. SIAM J. Mat. Anal. Appl. 38, 1454–1485 (2017)

    2017

  42. [50]

    PloS One 19, e0315,762 (2024)

    Viet Cuong, D., Lali´ c, B., Petri´ c, M., Thanh Binh, N., Roantree, M.: Adapting physics-informed neural networks to improve ODE optimization in mosquito populatio n dynamics. PloS One 19, e0315,762 (2024). doi:10.1371/journal.pone.0315762

    2024 doi

  43. [51]

    Weng, Y ., Zhou, D.: Multiscale physics-informed neural networks for stiff chemical kinetics. J. Phys. Chem. A 126, 8534–8543 (2022)

    2022

  44. [52]

    Williams, C., Seeger, M.: Using the Nystr¨ om method to sp eed up kernel machines. Adv. Neural. Inf. Process. Syst. 13, 682–688 (2001)

    2001

  45. [53]

    Woodruff, D.P .: Sketching as a tool for numerical linear algebra. Found. Trends. Theor. Comput. Sci. 10, 1–157 (2014)

    2014

  46. [54]

    Woolfe, F., Liberty, E., Rokhlin, V ., Tygert, M.: A fast r andomized algorithm for the approximation of matrices. Appl. Comput. Harmon. Anal. 25, 335–366 (2008)

    2008

  47. [55]

    Wu, C., Feng, X., He, Y ., Qian, L.: A second-order Strang s plitting scheme with exponential in- tegrating factor for the Allen–Cahn equation with logarith mic Flory–Huggins potential. Commun. Nonlinear Sci. Numer. 117, No. 106,983 (2023). doi:10.1016/j.cnsns.2022.106983

    2023

  48. [56]

    Zhang, H., Bi, Y ., Lavaei, J.: General low-rank matrix op timization: Geometric analysis and sharper bounds. Adv. Neural. Inf. Process. Syst. 34, 27,369–27,380 (2021)

    2021

  49. [57]

    Zhao, Y .L., Gu, X.M.: An adaptive low-rank splitting app roach for the extended Fisher–Kolmogorov equation. J. Comput. Phys. 506, No. 112,925 (2024). doi:10.1016/j.jcp.2024.112925

    2024

  50. [58]

    Zhao, Y .L., Ostermann, A., Gu, X.M.: A low-rank Lie-Trot ter splitting approach for nonlin- ear fractional complex Ginzburg-Landau equations. J. Comp ut. Phys. 446, No. 110,652 (2021). doi:10.1016/j.jcp.2021.110652

    2021

  51. [59]

    Zhou, X.Y ., Li, D.: Continuous-time mean-variance port folio selection: A stochastic LQ framework. Appl. Math. Optim. 42, 19–33 (2000)

    2000

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.