Pith. sign in

REVIEW 3 major objections 5 minor 51 references

This paper establishes a practical, online scheme — the model reference adaptive system — that recovers unknown spatially dependent parameters in nonlinear parabolic PDEs from time series data, with stable reconstructions even under heavy n

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-03 00:56 UTC pith:D3NT4AP7

load-bearing objection A credible first numerical implementation of the authors' MRAS, with public code and careful benchmark derivations, but the 'verified assumptions' claim overreaches because the convergence theorem does not cover the linear-parameter cases or the noisy-data runs. the 3 major comments →

arxiv 2602.10920 v2 pith:D3NT4AP7 submitted 2026-02-11 math.OC math.APmath.DS

Data assimilation via model reference adaptation for linear and nonlinear dynamical systems

Benedikt Kaltenbach , Christian Aarset , Tram Thi Ngoc Nguyen This is my paper
classification math.OC math.APmath.DS MSC 65M3265J2235R30
keywords data assimilationmodel reference adaptive systemonline parameter identificationnonlinear parabolic PDEsreal-time inversioncoefficient identificationsemi-implicit time steppingfinite element method
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper turns a theoretically developed model reference adaptive system (MRAS) into a working numerical method for online parameter identification in partial differential equations. The method couples a modified state equation, which becomes linear in the state by feeding observed data into the nonlinearity, with a parameter evolution law driven by model–data residuals and an adjoint derivative term. The authors verify the analytic conditions needed for exponential convergence and demonstrate stable reconstructions of unknown spatially dependent coefficients in four parabolic benchmark problems, including one with 20% measurement noise. A sympathetic reader would take away that MRAS offers a deterministic, real-time alternative to statistical data assimilation for a wide class of nonlinear inverse problems.

Core claim

On the paper's own terms, the central claim is that the MRAS (4) — two coupled evolution equations for the parameter q and state u, with the state correction operator C(||q||_H) chosen coercive per Assumption (A3) — provides numerically stable, online reconstructions of spatially varying parameters in parabolic PDEs, even when the model is nonlinear in both state and parameter. For each of four examples, the authors verify the required Lipschitz and coercivity conditions (in the nonlinear parameter cases, using the positivity and uniform boundedness of the true state) and implement a semi-implicit Euler / finite element scheme whose alternating prediction-correction structure matches the the

What carries the argument

The central object is the MRAS system (4): D_t q + σ[D_t z + f(q,z) − g] − f'_q(˜q,z)*(u−z) = 0 and D_t u + f(q,z) + C(||q||_H)(u−z) = g, with σ = 0 for models linear in q and σ = 1 otherwise. Its key feature is that the nonlinearity in the state acts on the observed data z, so the state equation is affine linear in u, while the parameter equation remains nonlinear in q and is driven by the residual between model prediction and data plus an adjoint feedback term. The operator C(||q||_H) is chosen to satisfy coercivity and boundedness (A3), producing a closed error system with exponential decay.

Load-bearing premise

The convergence proofs rest on the measured state staying strictly positive and uniformly bounded (inequality (22)) to obtain coercivity; for the two linear-parameter examples the paper does not actually verify coercivity and instead leans on numerical success.

What would settle it

Run the nonlinear potential problem with data that crosses zero while all other settings are unchanged; if the parameter estimate still converges, the positivity assumption is not load-bearing, but if it diverges or oscillates, the proof's coercivity step is essential. Separately, on the Darcy problem, monitor the parameter error over many steps without the u∆a modification; a blow-up would contradict the paper's empirical claim of stable convergence.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Calibrated on full-state time series, the MRAS recovers both the unknown parameter and the state online, so no offline inversion is needed.
  • Because the nonlinearity is evaluated on data, the state solve at each step is linear; this keeps per-step cost comparable to solving a linear parabolic PDE.
  • Convergence is exponential under Assumption 1, with a rate governed by the coercivity constant of the model and the feedback operator C.
  • As demonstrated up to 20% relative noise, the method tolerates substantial measurement error without blowing up, making it a candidate for real-time monitoring.

Where Pith is reading between the lines

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

  • If the positivity condition (22) is really necessary, the method would fail for data that changes sign; a targeted experiment with sign-changing data would delineate the true domain of applicability.
  • The scheme's reliance on full-state observations could be relaxed to trace or partial observations using observability arguments, but the convergence proof would need substantial extension, as the paper notes.
  • Because the state equation is linear, the method may combine naturally with reduced-order models or GPU solvers for real-time weather or reservoir forecasting, where 4D-Var is currently the standard but costly.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. This paper presents a numerical implementation of the model reference adaptive system (MRAS) proposed in [1] for online parameter identification in parabolic PDEs. A semi-implicit finite-element time-stepping scheme is introduced for the coupled state-parameter system, and four benchmark problems are studied: Darcy flow, Fisher–KPP, a nonlinear potential equation, and a modified Allen–Cahn equation. For each case the authors derive the MRAS components, discuss assumptions, and report reconstructions from synthetic data, including noisy data. The main claim is that the method is a reliable, versatile tool for real-time data assimilation, with verified assumptions in all cases.

Significance. If the stated guarantees held, this would be a valuable practical bridge between the abstract convergence theory of [1] and numerical online inversion for PDEs with nonlinear parameter dependence. The paper ships reproducible public code, treats four nontrivial benchmark problems, and demonstrates empirically stable reconstructions, including at 20% noise in the Allen–Cahn case. However, as detailed below, the convergence theory does not cover several of the reported experiments: coercivity is explicitly not verified for the linear-parameter cases, and for the nonlinear cases the verification relies on exact noise-free data and pointwise positivity. The numerical results are still informative as empirical evidence, but the paper's central claim of 'verified assumptions' across all runs needs substantial qualification.

major comments (3)
  1. [Abstract; §4.1; §5.1] The abstract claims 'verified assumptions' for all cases, but for the Darcy problem §4.1 states that coercivity (A2) 'will require to perturb the underlying equation (11) by adding u∆a' and that this modification is not implemented. The same gap applies to the Fisher–KPP case in §5.1, which is declared analogous. Since Proposition 1 requires (A2), the convergence theorem does not apply to the numerical experiments in Sections 4 and 5. This should be stated explicitly in the abstract and in the relevant sections; at present the claim is inaccurate.
  2. [§6.1, Eq. (25); §7.1] For the nonlinear potential and Allen–Cahn cases, coercivity (A2) is proved by identifying z with the true state u† and using the positivity lower bound (22). The constant C_coe = z in (25) and the coercive operator in (24) depend on z being pointwise positive. In the noisy-data experiments of §6.2 (5% noise) and §7.2 (5–20% noise) the data z are corrupted exact states and need not be positive, so the monotonicity argument and the coercivity constant are not available. Thus Proposition 1 does not justify the noisy nonlinear reconstructions. Either the theory must be extended to noisy data (e.g., by a robust positivity condition or a regularized data model) or the noise-free theoretical scope must be stated explicitly.
  3. [§3, Eqs. (8)–(9)] The semi-implicit scheme is introduced without any convergence analysis connecting it to the continuous MRAS (4). Equation (9) also uses D_t z^{n+1}, i.e., a numerical derivative of the data, which is not well-posed for noisy data without regularization. Consequently, even where Assumption 1 is satisfied, Proposition 1 governs the continuous MRAS, not the discretized system actually simulated. The paper should either prove consistency/stability/convergence of the scheme or clearly label the numerical study as an empirical investigation that is not covered by the cited theorem.
minor comments (5)
  1. [§5 title] The equation is standardly called Fisher–KPP (Kolmogorov–Petrovsky–Piskunov), not Fisher–KKP. The typo recurs throughout Section 5.
  2. [Eq. (4) and Eq. (9)] The term D_t z appears in the continuous MRAS and in the discrete update. For noisy data this term is not well-defined; a comment on how the time derivative of data is computed in the experiments would improve reproducibility.
  3. [§6.2, Table 4] The domain is written as B_0(π) in Table 4 while the text uses B_π(0). Please make the notation consistent.
  4. [§4.2, Fig. 4] The text says the initial parameter a0 is the indicator of a ball of radius 0.42, but the figure shows a0 with values 0 and 1 on a mesh; clarify whether the discontinuity is represented via the piecewise-constant parameter space or by a smoothed indicator.
  5. [§2.3] Proposition 1 is quoted from [1] with proof omitted. A short remark that the proof is machine-checked or otherwise verified would help, but at minimum the statement 'Proof. [1, Proposition 2.1]' should be flagged as a citation rather than a proof.

Circularity Check

0 steps flagged

No circular derivation: the MRAS parameter reconstructions are genuine outputs of forward-simulated data, though the abstract's 'verified assumptions' overstates coverage because Assumption (A2) is explicitly not established for the linear-parameter cases and is not robust for noisy nonlinear data.

full rationale

The paper's derivation chain is not circular. The MRAS system (4) is transparently imported from the authors' prior work [1], and the convergence result Proposition 1 is quoted from [1, Proposition 2.1]; this is a normal citation of a published theoretical result, not a hidden re-use of the present paper's own outputs. The numerical benchmarks are genuine synthetic tests: ground-truth parameters a†, c† are fixed, the forward PDE is solved to generate u†, and the MRAS is then run on data z derived from u†. The parameter updates (e.g., Eq. (13): ∫ a^{n+1} s dx = ∫ a^n s dx + Δt ∫ ∇z^n·∇(u^n-z^n) s dx) are driven by model-data residuals, and the reported reconstructions are not fitted constants or predetermined by the inputs. No quantity called a 'prediction' is equal by construction to a fitted parameter. The self-citation to [1] supplies the convergence theory, but the independent forward-model experiments provide separate empirical evidence, so the citation is not load-bearing circularity. The main legitimate concerns are correctness gaps, not circularity: Section 4.1 explicitly states that coercivity (A2) 'will require to perturb the underlying equation (11) by adding uΔa, as discussed in [1, Section 3.2]' and that this modification is not implemented; the Fisher–KPP case does not verify (A2) at all; and the nonlinear-case proof of (A2) relies on positivity (22) of u†, which noisy data (5–20% noise) need not satisfy. Also, the implemented feedback constant C_cn in Corollary 3 contains ‖c†‖_{L^2}, so the algorithm uses a norm of the ground truth, weakening the 'no prior information' claim. These are limitations in matching the stated assumptions, but they do not make the derivation equivalent to its inputs. No circular step can be exhibited, so the circularity score is 0.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 0 invented entities

The paper introduces no new physical entities. Its load-bearing assumptions are the positivity of the true state (for coercivity in nonlinear cases), the unproven discrete stability, and the correctness of the self-cited continuous convergence theory from [1].

free parameters (4)
  • time step Δt = 0.001
    Discretization parameter chosen for stability; not fitted to data.
  • mesh size h_max = 0.04 to 0.1 depending on example
    Spatial discretization parameter; chosen for accuracy.
  • feedback constants M and C in C(||c||) = M = 1/C, C = 4(3/5)^2
    Chosen in Remark 2 to satisfy Assumption A3; any positive M and C≥1 work, so not fitted to data.
  • Sobolev embedding constant C_{H^1→L^6} = not specified
    Appears in the feedback operator; value not given in the paper, needed to reproduce code.
axioms (4)
  • domain assumption The forward PDE (1) is uniquely solvable for u† at the true parameter q†.
    Assumed in Section 2.1 without proof; standard well-posedness assumption for parabolic PDEs.
  • domain assumption The true state u† is positive and uniformly bounded: 0 < z ≤ u† ≤ z̄ a.e.
    Used to prove coercivity (A2) in Propositions 4 and 5; proven in [1] via maximum principle, but not in this paper.
  • ad hoc to paper Convergence results from [1, Proposition 2.1] quoted as Proposition 1 are correct.
    The proof is not reproduced; the paper's theoretical grounding is entirely self-cited.
  • ad hoc to paper The semi-implicit discrete scheme is a stable and convergent approximation of the continuous MRAS.
    No discrete analysis is provided; numerical stability is only empirical.

pith-pipeline@v1.3.0-alltime-deepseek · 21180 in / 18383 out tokens · 160882 ms · 2026-08-03T00:56:17.230693+00:00 · methodology

0 comments
read the original abstract

We address data assimilation for linear and nonlinear dynamical systems via the so-called model reference adaptive system. Continuing our theoretical developments, we deliver the first practical implementation of this approach for online parameter identification with time series data. Our semi-implicit scheme couples a modified state equation with a parameter evolution law that is driven by model-data residuals. We demonstrate four benchmark problems of increasing complexity: the Darcy flow, the Fisher-KPP equation, a nonlinear potential equation and finally, an Allen-Cahn type equation. Across all cases, explicit model reference adaptive system construction, verified assumptions and numerically stable reconstructions underline our proposed method as a reliable, versatile tool for data assimilation and real-time inversion.

Figures

Figures reproduced from arXiv: 2602.10920 by Benedikt Kaltenbach, Christian Aarset, Tram Thi Ngoc Nguyen.

Figure 1
Figure 1. Figure 1: Parameters in benchmark examples, including Darcy flow, Fisher [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Schematic of data assimilation for the state [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: MRAS workflow for dynamic update laws 4 Darcy flow: the linear a-problem As our first case study, we consider the Darcy flow with homogeneous Dirich￾let boundary and unknown spatially dependent diffusion a defined over the unit square. That is, Dtu − ∇ · (a∇u) = g in I × Ω := [0,∞) × (0, 1)2 , u|∂Ω = 0 in I, u(t = 0) = u0 in Ω. (11) 11 [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Darcy flow. Visualization of evolution of the state [PITH_FULL_IMAGE:figures/full_fig_p015_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Fisher-KPP. Visualization of evolution of the state [PITH_FULL_IMAGE:figures/full_fig_p018_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Fisher-KPP. Error field a(T) − a † for MRAS output a given clean data (left) and 3% noisy data (right). Proposition 4. For the nonlinear potential problem (21) with unknown potential c and data z, the MRAS (4) takes the form Dtc + σ  Dtz − ∆z + cz + c|c| 2 3 z − g  = z  1 + 5 3 |c˜| 2 3  (u − z), Dtu − ∆z + cz + c|c| 2 3 z + C(∥q∥H)(u − z) = g, (c, u)(0) = (c0, u0), (23) where σ = 1, with the state spa… view at source ↗
Figure 7
Figure 7. Figure 7: Nonlinear potential. Evolution of the state [PITH_FULL_IMAGE:figures/full_fig_p025_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Nonlinear potential. Error field c(T) − c † resulted from MRAS ran with clean data (left) and data with 5% noise (right). Proof. Since (32) is a modification of (21) with the reaction law cu3 instead of cu. As this is nonlinearity in the state, one need only change cz to cz3 in f(c, z) whenever it appears. Also by this reason, Lipschitz continuity w. r. t parameter remains unchanged, resulting in the same … view at source ↗
Figure 9
Figure 9. Figure 9: Modified Allen-Cahn equation. Visualization of evolution of the [PITH_FULL_IMAGE:figures/full_fig_p029_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Modified Allen-Cahn equation. Error field [PITH_FULL_IMAGE:figures/full_fig_p030_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Modified Allen-Cahn-equation. L 2 -Error plots for the potential c and the state u. 31 [PITH_FULL_IMAGE:figures/full_fig_p031_11.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

51 extracted references · 12 canonical work pages

  1. [1]

    A model reference adaptive system approach for nonlinear online parameter identification

    B. Kaltenbacher and T. T. N. Nguyen. “A model reference adaptive system approach for nonlinear online parameter identification”. In: Inverse Problems37.5 (Apr. 2021), p. 055006.doi:10.1088/1361- 6420/abf164

    doi:10.1088/1361- 2021

  2. [2]

    Ensemble Kalman methods with constraints

    D. J. Albers, P.-A. Blancquart, M. E. Levine, E. E. Seylabi, and A. Stu- art. “Ensemble Kalman methods with constraints”. In:Inverse Prob- lems35.9 (2019), p. 095007.doi:10.1088/1361-6420/ab1c09

    work page doi:10.1088/1361-6420/ab1c09 2019

  3. [3]

    Analysis of the Ensemble Kalman Filter for Inverse Problems

    C. Schillings and A. M. Stuart. “Analysis of the Ensemble Kalman Filter for Inverse Problems”. In:SIAM Journal on Numerical Analysis 55.3 (2017), pp. 1264–1290.doi:10.1137/16M105959X

    doi:10.1137/16m105959x 2017

  4. [4]

    A reduced basis Kalman filter for parametrized partial differential equations

    M. Dihlmann and B. Haasdonk. “A reduced basis Kalman filter for parametrized partial differential equations”. In:ESAIM: Control, Op- timisation and Calculus of Variations(2015).doi:10.1051/cocv/ 2015019

    doi:10.1051/cocv/ 2015

  5. [5]

    Variational Data Assimilation with a Learned Inverse Observation Operator

    T. Frerix et al. “Variational Data Assimilation with a Learned Inverse Observation Operator”. In:Proceedings of the 38th International Con- ference on Machine Learning. Proceedings of Machine Learning Re- search. 2021, pp. 3449–3458.url:https://proceedings.mlr.press/ v139/frerix21a.html

    2021

  6. [6]

    Variational data assimilation: theory and overview

    F. Rabier and Z. Liu. “Variational data assimilation: theory and overview”. In:ECMWF Seminar on Recent Developments in Data Assimilation for Atmosphere and Ocean. Seminar proceedings, 8–12 September 2003. ECMWF. Reading, UK, 2003.url:https://www.ecmwf.int/en/ elibrary/76079- variational- data- assimiltion- theory- and- overview

    2003

  7. [7]

    A global three-dimensional multivariate statistical in- terpolation scheme

    A. C. Lorenc. “A global three-dimensional multivariate statistical in- terpolation scheme”. In:Monthly Weather Review109.4 (1981), pp. 701– 721.doi:https : / / doi . org / 10 . 1175 / 1520 - 0493(1981 ) 109 % 3C0701:AGTDMS%3E2.0.CO;2

    1981

  8. [8]

    Analysis methods for numerical weather prediction

    A. C. Lorenc. “Analysis methods for numerical weather prediction”. In:Quarterly Journal of the Royal Meteorological Society112.474 (1986), pp. 1177–1194.doi:10.1002/qj.49711247414

    doi:10.1002/qj.49711247414 1986

  9. [9]

    Variational algorithms for analysis and assimilation of meteorological observations: theoretical aspects

    F.-X. Le Dimet and O. Talagrand. “Variational algorithms for analysis and assimilation of meteorological observations: theoretical aspects”. In:Tellus A38.2 (1986), pp. 97–110.doi:10.1111/j.1600- 0870. 1986.tb00459.x. 32

    doi:10.1111/j.1600- 1986

  10. [10]

    The use of adjoint equations to solve a variational adjustment problem with advective constraints

    J. M. Lewis and J. C. Derber. “The use of adjoint equations to solve a variational adjustment problem with advective constraints”. In:Tellus A:37.4 (1985), pp. 309–322.doi:10 . 1111 / j . 1600 - 0870 . 1985 . tb00430.x

    1985

  11. [11]

    Reich.Probabilistic Forecasting and Bayesian Data Assimilation

    S. Reich.Probabilistic Forecasting and Bayesian Data Assimilation. Cambridge University Press, 2015.isbn: 978-1107663916

    2015

  12. [12]

    Kaltenbacher, A

    B. Kaltenbacher, A. Neubauer, and O. Scherzer.Iterative Regulariza- tion Methods for Nonlinear Ill-posed Problems. de Gruyter, Berlin: Radon Series on Computational and Applied Mathematics, 2008.doi: https://doi.org/10.1515/9783110208276

    doi:10.1515/9783110208276 2008

  13. [13]

    Springer, 2011.doi:10.1007/978-1-4419-8474-6

    Kirsch.An Introduction to the Mathematical Theory of Inverse Prob- lems. Springer, 2011.doi:10.1007/978-1-4419-8474-6

    doi:10.1007/978-1-4419-8474-6 2011

  14. [15]

    Narendra and A

    K. Narendra and A. Annaswamy.Stable Adaptive Systems. Mineola, New York: Dover Publications, 2005.isbn: 978-0-13-839994-8

    2005

  15. [16]

    Ioannou and J

    P. Ioannou and J. Sun.Robust Adaptive Control. Englewood Cliffs: Prentice Hall, 1996.isbn: 978-0-13-439100-7

    1996

  16. [17]

    Sastry and M

    S. Sastry and M. Bodson.Adaptive Control: Stability, Convergence and Robustness. Mineola, New York: Dover Publications, 2011.isbn: 978-0-486-48202-6

    2011

  17. [18]

    On- Line Parameter Estimation for Infinite-Dimensional Dynamical Sys- tems

    J. Baumeister, W. Scondo, M. A. Demetriou, and I. G. Rosen. “On- Line Parameter Estimation for Infinite-Dimensional Dynamical Sys- tems”. In:SIAM Journal on Control and Optimization35.2 (1997), pp. 678–713.doi:10.1137/S0363012994270928

    work page doi:10.1137/s0363012994270928 1997

  18. [19]

    Online parameter identification in time-dependent differ- ential equations as a non-linear inverse problem

    P. K¨ ugler. “Online parameter identification in time-dependent differ- ential equations as a non-linear inverse problem”. In:European Jour- nal of Applied Mathematics19.5 (2008), pp. 479–506.doi:10.1017/ S0956792508007547

    2008

  19. [20]

    K¨ ugler

    P. K¨ ugler. “Online parameter identification without Ricatti-type equa- tions in a class of time-dependent partial differential equations: an extended state approach with potential to partial observations”. In: Inverse Problems26, 035004 (2010). Art. ID 035004.doi:10.1088/ 0266-5611/26/3/035004. 33

    2010

  20. [21]

    An online parameter identification method for time dependent partial differential equations

    R. Boiger and B. Kaltenbacher. “An online parameter identification method for time dependent partial differential equations”. In:Inverse Problems32.4 (Mar. 2016), p. 045006.doi:10.1088/0266-5611/32/ 4/045006

    work page doi:10.1088/0266-5611/32/ 2016

  21. [22]

    Bi-level iterative regularization for inverse problems in nonlinear PDEs

    T. T. N. Nguyen. “Bi-level iterative regularization for inverse problems in nonlinear PDEs”. In:Inverse Probl.40.4 (2024), p. 045020.doi: 10.1088/1361-6420/ad2905

    work page doi:10.1088/1361-6420/ad2905 2024

  22. [23]

    Sequential bi-level regularized inversion with ap- plication to hidden reaction law discovery

    T. T. N. Nguyen. “Sequential bi-level regularized inversion with ap- plication to hidden reaction law discovery”. In:Inverse Probl.41.6 (2025), p. 065015.doi:10.1088/1361-6420/addf73

    doi:10.1088/1361-6420/addf73 2025

  23. [24]

    The Tangential Cone Condition for Some Coefficient Identification Model Problems in Parabolic PDEs

    B. Kaltenbacher, T. T. N. Nguyen, and O. Scherzer. “The Tangential Cone Condition for Some Coefficient Identification Model Problems in Parabolic PDEs”. In:Time-dependent Problems in Imaging and Parameter Identification. Ed. by B. Kaltenbacher, T. Schuster, and A. Wald. Cham: Springer International Publishing, 2021, pp. 121–163. isbn: 978-3-030-57784-1.do...

    work page doi:10.1007/978-3-030-57784-1_5 2021

  24. [25]

    Roub ´ ıˇ cek.Nonlinear Partial Differential Equations with Applica- tions

    T. Roub ´ ıˇ cek.Nonlinear Partial Differential Equations with Applica- tions. Springer, 2013.isbn: 978-3034805124

    2013

  25. [26]

    NETGEN An advancing front 2D/3D-mesh generator based on abstract rules

    J. Sch¨ oberl. “NETGEN An advancing front 2D/3D-mesh generator based on abstract rules”. In:Computing and Visualization in Sci- ence1.1 (July 1997), pp. 41–52.issn: 1432-9360.doi:10 . 1007 / s007910050004

    1997

  26. [27]

    Pei.The Semi-implicit DLN Algorithm for the Navier Stokes Equa- tions

    W. Pei.The Semi-implicit DLN Algorithm for the Navier Stokes Equa- tions. 2023. arXiv:2306.02461 [math.NA]

    Pith/arXiv arXiv 2023

  27. [28]

    Zhao and Z

    Y. Zhao and Z. Zhou.Efficient bound preserving and asymptotic pre- serving semi-implicit schemes for the fast reaction-diffusion system

  28. [29]

    Netgen/NGSolve team.Dirichlet boundary conditions.https://docu. ngsolve . org / nightly / i - tutorials / unit - 1 . 3 - dirichlet / dirichlet.html. 2017

    2017

  29. [30]

    Preconditioning strategies for vectorial finite element linear systems arising from phase-field models for fracture mechanics

    M. Badri, G. Rastiello, and E. Foerster. “Preconditioning strategies for vectorial finite element linear systems arising from phase-field models for fracture mechanics”. In:Computer Methods in Applied Mechanics and Engineering373 (2021), p. 113472.issn: 0045-7825.doi:https: //doi.org/10.1016/j.cma.2020.113472

    arXiv 2021

  30. [31]

    B. M. Kaltenbach, C. Aarset, and T. T. N. Nguyen.Data assimila- tion via model reference adaptation for linear and nonlinear dynamical systems. Version V3. 2026.doi:10.25625/HWM9EJ. 34

    work page doi:10.25625/hwm9ej 2026

  31. [32]

    Iterative regularization of parameter identification problems by sequential quadratic programming meth- ods

    M. Burger and W. M¨ uhlhuber. “Iterative regularization of parameter identification problems by sequential quadratic programming meth- ods”. In:Inverse Problems18.4 (May 2002), p. 943.doi:10.1088/ 0266-5611/18/4/301

    2002

  32. [33]

    Numerical Approximation of an SQP- Type Method for Parameter Identification

    M. Burger and W. M¨ uhlhuber. “Numerical Approximation of an SQP- Type Method for Parameter Identification”. In:SIAM Journal on Nu- merical Analysis40.5 (2002), pp. 1775–1797.doi:10.1137/S0036142901389980

    work page doi:10.1137/s0036142901389980 2002

  33. [34]

    Preconditioned all-at-once methods for large, sparse parameter estimation problems

    E. Haber and U. M. Ascher. “Preconditioned all-at-once methods for large, sparse parameter estimation problems”. In:Inverse Problems 17.6 (Nov. 2001), p. 1847.doi:10.1088/0266-5611/17/6/319

    work page doi:10.1088/0266-5611/17/6/319 2001

  34. [35]

    Goal oriented adaptiv- ity in the IRGNM for parameter identification in PDEs: II. all-at-once formulations

    B. Kaltenbacher, A. Kirchner, and B. Vexler. “Goal oriented adaptiv- ity in the IRGNM for parameter identification in PDEs: II. all-at-once formulations”. In:Inverse Problems30.4 (Feb. 2014), p. 045002.doi: 10.1088/0266-5611/30/4/045002

    doi:10.1088/0266-5611/30/4/045002 2014

  35. [36]

    A penalty method for PDE- constrained optimization in inverse problems

    T. van Leeuwen and F. J. Herrmann. “A penalty method for PDE- constrained optimization in inverse problems”. In:Inverse Probl.32.1 (2016), p. 015007.url:http://stacks.iop.org/0266-5611/32/i= 1/a=015007

    2016

  36. [37]

    Landweber–Kaczmarz for parameter identification in time-dependent inverse problems: all-at-once vs reduced version

    T. T. N. Nguyen. “Landweber–Kaczmarz for parameter identification in time-dependent inverse problems: all-at-once vs reduced version”. In:Inverse Problems35.3 (2019), p. 035009.doi:10 . 1088 / 1361 - 6420/aaf9ba

    2019

  37. [38]

    Regularization Based on All-At-Once Formulations for Inverse Problems

    B. Kaltenbacher. “Regularization Based on All-At-Once Formulations for Inverse Problems”. In:SIAM Journal on Numerical Analysis54.4 (2016), pp. 2594–2618.doi:10.1137/16M1060984

    doi:10.1137/16m1060984 2016

  38. [39]

    Kaltenbacher, T

    B. Kaltenbacher, T. T. N. Nguyen, A. Wald, and T. Schuster.Parame- ter identification for the Landau-Lifshitz-Gilbert equation in Magnetic Partarticle Imaging. Springer, 2021, pp. 377–412.doi:10.1007/978- 3-030-57784-1_13

    doi:10.1007/978- 2021

  39. [40]

    Discretization of parameter identification in PDEs using neural networks

    B. Kaltenbacher and T. T. N. Nguyen. “Discretization of parameter identification in PDEs using neural networks”. In:Inverse Problems 38.12 (Nov. 2022), p. 124007.doi:10.1088/1361-6420/ac9c25

    doi:10.1088/1361-6420/ac9c25 2022

  40. [41]

    Learning-informed pa- rameter identification in nonlinear time-dependent PDEs

    C. Aarset, M. Holler, and T. T. N. Nguyen. “Learning-informed pa- rameter identification in nonlinear time-dependent PDEs”. In:Ap- plied Mathematics and Optimization88 (2023), 53 pp.doi:10.1007/ s00245-023-10044-y

    2023

  41. [42]

    Landau, L

    L. Landau, L. Pitaevskii, A. Kosevich, and E. Lifshitz.Theory of Elas- ticity: Volume 7. Course of theoretical physics. Elsevier Science, 1986. isbn: 9780750626330. 35

    1986

  42. [43]

    Darcy’s law and the field equations of the flow of un- derground fluids

    M. K. Hubbert. “Darcy’s law and the field equations of the flow of un- derground fluids”. In:International Association of Scientific Hydrol- ogy. Bulletin2.1 (1957), pp. 23–59.doi:10.1080/02626665709493062

    work page doi:10.1080/02626665709493062 1957

  43. [44]

    Takamoto et al.PDEBENCH: An Extensive Benchmark for Sci- entific Machine Learning

    M. Takamoto et al.PDEBENCH: An Extensive Benchmark for Sci- entific Machine Learning. 2024. arXiv:2210.07182 [cs.LG]

    Pith/arXiv arXiv 2024

  44. [45]

    Higher-order total variation approaches and generalisations

    K. Bredies and M. Holler. “Higher-order total variation approaches and generalisations”. In:Inverse Probl.36.12 (Dec. 2020), p. 123001. doi:10.1088/1361-6420/ab8f80

    work page doi:10.1088/1361-6420/ab8f80 2020

  45. [46]

    Revisiting the Fisher–Kolmogorov–Petrovsky–Piskunov equation to interpret the spreading–extinction dichotomy

    M. El-Hachem, S. W. McCue, W. Jin, Y. Du, and M. J. Simpson. “Revisiting the Fisher–Kolmogorov–Petrovsky–Piskunov equation to interpret the spreading–extinction dichotomy”. In:Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 475.2229 (Sept. 2019).doi:https://doi.org/10.1098/rspa.2019. 0378

    doi:10.1098/rspa.2019 2019

  46. [47]

    Investigation of a structured Fisher’s equation with applications in biochemistry

    J. T. Nardini and D. M. Bortz. “Investigation of a structured Fisher’s equation with applications in biochemistry”. In:SIAM Journal on Applied Mathematics78.3 (2018), pp. 1712–1736.doi:https://doi. org/10.1137/16M1108546

    work page doi:10.1137/16m1108546 2018

  47. [48]

    C. V. Pao.Nonlinear Parabolic and Elliptic Equations. New York and London: Plenum Press, 1992.isbn: 9780306443435

    1992

  48. [49]

    Ground state structures in ordered binary alloys with second neighbor interactions

    S. Allen and J. Cahn. “Ground state structures in ordered binary alloys with second neighbor interactions”. In:Acta Metallurgica20.3 (1972), pp. 423–433.issn: 0001-6160.doi:https://doi.org/10.1016/0001- 6160(72)90037-5

    doi:10.1016/0001- 1972

  49. [50]

    B. H. Gilding and R. Kersner.Travelling Waves in Nonlinear Diffusion- Convection Reaction. Switzerland: Springer Basel AG, 2004.doi:https: //doi.org/10.1007/978-3-0348-7964-4

    work page doi:10.1007/978-3-0348-7964-4 2004

  50. [51]

    The extended adjoint state and nonlinearity in correlation-based passive imaging

    T. T. N. Nguyen. “The extended adjoint state and nonlinearity in correlation-based passive imaging” (2025). arXiv:2504.16797 [math.NA]. 36

    Pith/arXiv arXiv 2025

  51. [2024]

    arXiv:2404.18463 [math.NA]

    Pith/arXiv arXiv