Pith. sign in

REVIEW 3 major objections 5 minor 55 references

Reduced Order Models and Conditional Expectation -- Analysing Parametric Low-Order Approximations

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A parametric reduced-order model built by least-squares projection is a conditional expectation, so every such ROM is a Bayesian update.

desk verdict A clear, honest synthesis paper that re-labels least-squares ROMs as conditional expectations; the label is defensible but largely definitional. read the letter →

arxiv 2412.19836 v2 pith:GQVGNALD submitted 2024-12-22 cs.LG cs.NAmath.NA

Hermann G. Matthies This is my paper · ORCID
classification cs.LGcs.NAmath.NA MSC 35B3035R3035R6041A4541A6360G2060G6060H15
keywords reducedordermodelsconditionalexpectationBayesianupdatingKarhunen-Loèveexpansionuncertaintyquantificationmachinelearningleastsquares
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

The paper argues that the standard way of building parametric reduced-order models—projecting the full model's solution onto a low-dimensional subspace by least squares—is mathematically the same operation as computing a conditional expectation of a random variable. Concretely, it shows that a ROM $r_a(\mu)$ computed this way satisfies $r_a(\mu) = E_g(r(\mu)|z)$, so the reduced state is the best least-squares prediction of the full state given the available parameter-dependent observations. If that identification is right, then proper orthogonal decomposition, Karhunen-Loève expansions, reduced-basis methods, Gaussian process emulation, low-rank tensor approximations, and much of machine learning are all instances of Bayesian updating with a mean-square-error loss. The payoff is practical: one can add the modeling, discretization, numerical, and reduction errors that are normally ignored when a ROM is built, and treat them as additional random variables in the same least-squares loss, producing reduced models that are optimal with respect to all these uncertainties at once.

What carries the argument

The mechanism that carries the argument is the encoding of the parametric family $\{r(p)\}$ into a linear map $R: U \to Q$ between Hilbert spaces, combined with the variational definition of conditional expectation as a least-squares projection. The linear map turns an unstructured parameter set into a vector space, so the correlation $C_U = R^\dagger R$ has a singular value decomposition whose truncation—the Karhunen-Loève expansion—gives the ROM; the conditional expectation, defined as the minimizer of $\Psi_x(\chi) = \|x - \chi\|^2$ over a subspace of functions of the observations, is the same least-squares operation with the same Galerkin/Pythagoras structure. Identifying the training/ROM sample space with the conditioning $\sigma$-algebra produces the identity $r_a(\mu) = E_g(r(\mu)|z)$, which is the bridge between model reduction and Bayesian updating.

What would settle it

Take a parametric problem where the map $p \mapsto r(p)$ is multi-valued, for instance a pitchfork bifurcation, and attempt to construct the ROM through Eq. (46); the conditional expectation returns a single averaged state that matches none of the branches, so the identity $r_a(\mu)=E_g(r(\mu)|z)$ fails exactly where the single-valued assumption breaks.

Watch

Extended reading notes

Core claim

The central claim is that a parametric reduced-order model obtained by least-squares projection can be identified with a conditional expectation (CEX). Encoding the parametric map $p \mapsto r(p)$ as a linear map $R$ and truncating its singular value expansion gives a ROM $r_a(\mu)$; the paper shows, through the variational characterization of conditional expectation as a least-squares projection onto a subspace of functions of the observed variable $z$, that Eq. (46) holds: $r_a(\mu) = E_g(r(\mu)|z)$. In the author's formulation, every such ROM 'may be seen as the CEX of $r(\mu)$', meaning the reduced model is the best mean-square prediction of the full state given the information $z$ under the chosen probability measure. This makes POD, KLE, RBM, Gaussian process emulation, low-rank tensor ROMs, and least-squares trained neural networks instances of one common Bayesian updating operation, and opens the way to include all the errors in the modeling chain as random variables in the loss.

Load-bearing premise

The construction assumes the parametric map $p \mapsto r(p)$ is single-valued for every $p$, so each parameter value has exactly one state; if a parameter value admits multiple solutions, the linear map $R$, the correlation $C_U = R^\dagger R$, and the conditional-expectation identity are not defined.

Editorial extensions

If this is right

  • Any ROM built from a least-squares projection—POD, KLE, reduced basis, Gaussian process emulation, low-rank tensor approximations—can be treated as an approximate Bayesian update, so filtering, sequential updating, and posterior error estimates apply to it.
  • The modeling, discretization, numerical, and reduction errors that are usually ignored when building a ROM can be included as random variables in the same squared-distance loss, making the reduced model optimal against total uncertainty rather than just the FOM-to-ROM error.
  • The identification gives a common vocabulary for reduced-order modeling and machine learning: training a surrogate by minimizing mean-square error on samples is an approximate conditional expectation, so error analyses and uncertainty tools transfer between the two fields.
  • Since any factorization of $C_U = R^\dagger R$ yields a different linear re-parametrisation with the same SVD structure, the conditional-expectation interpretation is invariant across the many ROM algorithms, not tied to one particular expansion.

Reading between the lines

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

  • A direct testable extension: if the CEX view is right, a ROM trained on samples drawn from one parameter distribution and then evaluated on another will be suboptimal exactly by the mismatch between the conditioning measure and the target measure; reweighting the loss to the target distribution should recover the optimal ROM on a standard parametric PDE benchmark.
  • For bifurcation problems, the conditional-expectation ROM would produce an averaged state between branches; a decision-oriented extension the paper does not develop would replace the mean-square loss with a risk-sensitive or multi-modal loss while keeping the projection-geometry viewpoint.
  • The Fréchet-mean generalisation suggests that non-Hilbertian losses such as robust or Wasserstein distances could replace mean-square error in ROM construction, but the paper only notes this possibility and does not analyse it.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes a unifying probabilistic interpretation of parametric reduced-order models (ROMs): any ROM obtained by a least-squares projection can be seen as a conditional expectation (CEX) of the full-order parametric map. Section 2 recalls the encoding of parametric objects r(p) into a linear map R, leading to RKHS, correlation operators, SVD/KLE expansions, and reduced representations. Section 3 reviews the variational definition of conditional expectation as an orthogonal projection and its role in Bayesian updating and filtering. Section 4 introduces a 'generalized conditional expectation' Eg defined as the squared-loss minimizer over a user-chosen manifold Sg(z), then claims that standard ROM constructions—GPE/Kriging, reduced basis methods, POD, truncated SVD, and low-rank tensor approximations—are instances of Eq. (46), ra(μ)=Eg(r(μ)|z). The paper also discusses how modeling, discretization, numerical, and reduction errors could be incorporated as additional random variables in the loss. The concluding sections acknowledge that the parametric map is assumed single-valued and that multiple-solution/bifurcation cases are deferred to future work.

Significance. If the central identification were a theorem, it would provide a genuinely useful bridge between model reduction and Bayesian inference, with the potential to justify probabilistic error models in ROMs and to motivate non-quadratic loss functions. The paper's background material is sound and competently assembled: the linear-map encoding, RKHS construction, correlation operator, SVD/KLE expansion, and the variational characterization of conditional expectation are standard and correctly stated. The concrete illustrations for GPE, RBM, POD, and low-rank tensors are instructive and show the broad intended scope. However, the load-bearing claim, Eq. (46), is not substantiated as stated; it is a definitional re-labeling rather than a proven equivalence. The paper would be acceptable as an interpretive/survey contribution if that status were made explicit, but in its current form it overstates the strength of the connection between standard ROM projection and Kolmogorov conditional expectation.

major comments (3)
  1. [Section 4.2, Eqs. (44)–(46)] The central identification ra(μ)=Eg(r(μ)|z) is true by stipulation rather than by proof. In Eq. (44), Eg(x|z) is defined as the argmin of the squared-loss functional Ψx over an arbitrary manifold Sg(z); in Eq. (45) the ROM loss is the same functional; hence Eq. (46) restates the definition of the ROM as a minimization over its own approximation manifold. This does not establish that the ROM is a conditional expectation in the Kolmogorov sense, because no construction is given of a random variable z and a σ-algebra σ(z) such that Sg(z)=L²(Ω,σ(z)). To make the claim substantive, the paper must either prove that such a z exists for the ROM methods discussed, or explicitly redefine Eg as a new, non-Kolmogorov object and then avoid calling the result a 'Bayesian update'.
  2. [Section 4.3, 4.4; Section 4.2, paragraph after Eq. (46)] For projection-based ROMs such as POD, RBM, and truncated SVD, no observation z is specified. A genuine conditional expectation is always a function of the conditioning σ-algebra; a projection onto an arbitrary finite-dimensional subspace of L² is generally not realizable as conditioning on any coarser observation. Without exhibiting z (or showing that the chosen subspace equals L²(Ω,σ(z)) for some z), the assertion that every least-squares ROM 'may be seen as the CEX of r(μ)' is an analogy, not a mathematical equivalence. The paper should either provide such a construction for representative cases or temper the claim to a formal analogy, which is still of expository value.
  3. [Section 4.4, Eq. (54)] The Fréchet-type extension in Eq. (54) is asserted without specifying the norm ∥·∥_L on L(Q,R^N). The accompanying text states that this is 'not a Hilbert norm', but no precise definition of the operator norm or of the underlying space is given, and no existence or uniqueness argument is provided for the argmin over the non-subspace manifold Sn,SVD(Zn). Since Eq. (54) is presented as a CEX analogue, this gap is load-bearing for the claim that the truncated SVD is an instance of the framework; either the norm should be specified (with existence/uniqueness addressed) or the statement should be clearly marked as heuristic.
minor comments (5)
  1. [Section 2.6, Eq. (24) and surrounding text] The notation is inconsistent between R in Eq. (20) and RT in subsequent equations (e.g., Eq. (24) defines Ra via ra(p)^T w, while earlier the map is called R). The paper should standardize the notation for the linear map and its transpose.
  2. [Section 4.3, Eq. (49)] The displayed formula κN(μ1,μ2)=rN(μ1)^T rN(μ1) should presumably be rN(μ1)^T rN(μ2); otherwise the kernel is not symmetric in its two arguments.
  3. [Section 4.3, before Eq. (51)] The dimension of KGPE is stated as R^{m×n}, but it should be R^{m×m}; also g(μ) is an m-vector so Eq. (51) is consistent with an m×m matrix.
  4. [Throughout] There are several typos and missing words: 'tey' for 'they' in Example 1 of Section 2, 'in in' in Section 2.6, 'te' for 'the' near the end of Section 3, and 'by by' in Section 3. A careful proofreading pass is needed.
  5. [Section 4.4, Eqs. (54)–(56)] The notation Et,F in Eq. (54) and Et in Eqs. (55), (56), (58), (59) is not consistently defined or explained; in particular, what the subscript F refers to is unclear. Please define all expectation operators explicitly before using them in the minimization statements.

Circularity Check

2 steps flagged · score 7.0 of 10

Eq. (46) identifies the ROM with a 'generalized conditional expectation' defined as the same least-squares argmin, making the central claim true by definition rather than by derivation.

  1. self definitional [Section 4.2, Eqs. (44)-(46)]
    "Eg(x|z) = arg min χ∈Sg(z) Ψx(χ), (44) ... ra(µ) = Eg(r(µ)|z). (46)"

    Eq. (44) defines the generalized conditional expectation Eg(x|z) as the minimizer of the squared-loss functional Ψx over an arbitrary, user-chosen manifold Sg(z). The ROM ra(µ) is, by Eq. (23) and the surrounding text, itself obtained by a least-squares minimization over a chosen subspace or manifold. Hence Eq. (46) holds by construction: one may take Sg(z) to be the ROM's approximation manifold, and the minimization defining ra is exactly the minimization defining Eg. The paper supplies no argument that Sg(z) equals L2(σ(z)) for some observation z, which would be needed to identify the ROM with a genuine Kolmogorov conditional expectation; absent that, Eq. (46) is a definitional reformulation rather than an independent theorem.

  2. renaming known result [Section 4.4, Eq. (56)]
    "The POD basis is again given as a CEX: V k,POD = Et(V|Zn) = arg min A∈Sn,POD(Zn) ΨZN (A)∈S n,POD(Zn) (56)"

    The manifold Sn,POD(Zn) is defined as {A = VV^T Zn | V ∈ Vk(RN)}, which is precisely the set of rank-k orthogonal projections of the snapshot matrix. Since Et is defined as the argmin over that manifold, Eq. (56) restates the defining optimization problem of POD; calling the result a conditional expectation relabels the construction without adding new content. This is the same definitional move as Eq. (46), applied to a specific known method.

full rationale

The paper's linear-algebra core (Section 2, SVD/KLE, the GPE formulas) is mostly standard mathematics and is not circular; the self-citations to [35,36] reproduce the needed arguments in the text and are not load-bearing in a way that forces the conclusion. The circularity is concentrated in the paper's central interpretive step: Eq. (44) defines a 'generalized conditional expectation' so broadly that any least-squares projection qualifies, and Eq. (46) then identifies the ROM with that object. The identification is a definitional reformulation, not an independent derivation, and the same pattern recurs when POD, KLE, RBM, and low-rank tensor methods are each shown to be 'given as a CEX' by choosing the manifold to be the method's own search set. The paper is honest about its interpretive character ('may be seen as', 'practically any least-squares procedure'), but Eq. (46) is still presented as the main result, and it reduces by construction. Score 7 reflects a central claim that is forced by definition, with independent supporting material elsewhere.

Assumptions & free parameters 2 free parameters · 5 assumptions · 1 invented entities

The central claim rests on standard Hilbert-space projection theory plus the author's prior linear-map framework. There are no fitted data-driven parameters, but the choice of Q inner product, the assumed error independence, the single-valued map condition, and the known covariance kernel are all unverified modeling inputs. No new physical entities are introduced; the only new mathematical object is the generalized conditional expectation.

free parameters (2)
  • Inner product / measure on the parameter space Q = No fitted value; chosen by the user.
    Eq. (13) defines the correlation CU = R†R using the Q inner product, and Eq. (47) uses a measure rho on the parameter space. Different choices produce different SVD bases and different ROMs, so the claimed conditional-expectation interpretation depends on this hand-chosen input.
  • GPE covariance kernel kappa_s = Assumed known or chosen by hand, not fitted.
    Section 4.3 says that in practice the covariance function is often not known and is assumed in a computationally convenient form. The Kriging predictor in Eq. (53) is completely determined by this kernel.
assumptions (5)
  • domain assumption For every parameter p, the equation F(r(p),p)=0 has a unique solution r(p), i.e. the parametric map is single-valued.
    Invoked in Section 2.1 via the implicit function theorem and restated in the Conclusions. If false, as in bifurcation problems, the map R in Eq. (7), the correlation CU, and the conditional-expectation identification in Eq. (46) are not defined.
  • standard math The spaces U and Q are Hilbert spaces, and the spectra used in the main representation are discrete.
    Sections 2.1 and 2.5 assume Hilbert spaces and, for the main expansion in Eq. (14), a discrete spectrum with positive eigenvalues; the continuous-spectrum case is deferred to Appendix A.
  • domain assumption The parameter set P is replaced by a vector subspace Q of R^P via duality, and the inner product on Q defines what is important.
    Section 2.5 uses the Q inner product to define CU and hence the SVD basis. This is a modeling decision that fixes which parameter directions matter, and it is not derived from data or external constraints.
  • domain assumption Modeling uncertainties on Omega_M and numerical uncertainties on Omega_N are independent.
    Section 4.1 states it is not unreasonable to assume this independence and then uses Fubini's theorem to factor the total expectation. If the error sources are dependent, the factorized expectation Eg = EN ○ EM does not hold.
  • domain assumption In the Gaussian process emulation example, the parametric field is a zero-mean Gaussian process with known covariance.
    Section 4.3 assumes rs(mu) is a Gaussian process with known kernel kappa_s to derive the Kriging predictor in Eq. (53). This assumption is standard for GPE but is an additional modeling input, not a consequence of the ROM framework.
invented entities (1)
  • Generalized conditional expectation Eg in Eq. (44)
    purpose: To make every least-squares ROM an instance of conditional expectation by allowing an arbitrary user-chosen manifold Sg(z) instead of the classical L2(Omega,B) subspace.
    Defined in Section 4.2, not derived from external benchmarks or a theorem. The identification with ROMs follows directly from this definition, which is the main source of the circularity noted in the review.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Reduced Order Models and Conditional Expectation -- Analysing Parametric Low-Order Approximations." pith.science (2026). https://pith.science/paper/GQVGNALD

@misc{pith2026241219836,
  author       = {Pith},
  title        = {Pith review of: Reduced Order Models and Conditional Expectation -- Analysing Parametric Low-Order Approximations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GQVGNALD}},
  note         = {Machine review of arXiv:2412.19836}
}
read the original abstract

Systems may depend on parameters which one may control, or which serve to optimise the system, or are imposed externally, or they could be uncertain. This last case is taken as the ``Leitmotiv'' for the following. A reduced order model is produced from the full order model by some kind of projection onto a relatively low-dimensional manifold or subspace. The parameter dependent reduction process produces a function of the parameters into the manifold. One now wants to examine the relation between the full and the reduced state for all possible parameter values of interest. Similarly, in the field of machine learning, also a function of the parameter set into the image space of the machine learning model is learned on a training set of samples, typically minimising the mean-square error. This set may be seen as a sample from some probability distribution, and thus the training is an approximate computation of the expectation, giving an approximation to the conditional expectation, a special case of an Bayesian updating where the Bayesian loss function is the mean-square error. This offers the possibility of having a combined look at these methods, and also of introducing more general loss functions.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

55 extracted references · 30 canonical work pages

  1. [1]

    Batlle, Y

    P. Batlle, Y. Chen, B. Hosseini, H. Owhadi, and A. M. Stuart,Error Analysis of Kernel/GP Methods for Nonlinear and Parametric PDEs, arXiv:2305.04962 [math.NA], 2023, Available from: http: //arxiv.org/abs/2305.04962, arXiv:2305.04962

    arXiv 2023

  2. [2]

    Benner, S

    P. Benner, S. Gugercin, and K. Willcox,A survey of projection-based model reduction methods for parametric dynamical systems, SIAM Review57 (2015), 483–531, doi:10.1137/130932715

    doi:10.1137/130932715 2015

  3. [3]

    Benner, M

    P. Benner, M. Ohlberger, A. T. Patera, G. Rozza, and K. Urban (eds.),Model Reduction of Para- metrized Systems, MS&A — Modeling, Simulation & Applications, vol. 17, Berlin, Springer, 2017, doi:10.1007/978-3-319-58786-8

    doi:10.1007/978-3-319-58786-8 2017

  4. [4]

    Berlinet and C

    A. Berlinet and C. Thomas-Agnan,Reproducing kernel Hilbert spaces in probability and statistics, Springer, Berlin, 2004,doi:10.1007/978-1-4419-9096-9. ANALYSING PARAMETRIC ROMS AND CONDITIONAL EXPECTATION 23

    doi:10.1007/978-1-4419-9096-9 2004

  5. [5]

    Binev, A

    P. Binev, A. Cohen, W. Dahmen, R. DeVore, G. Petrova, and P. Wojtaszczyk,Data assimilation in reduced modeling, SIAM/ASA Journal of Uncertainty Quantification5 (2017), no. 1, 1–29,doi: 10.1137/15M1025384

    work page doi:10.1137/15m1025384 2017

  6. [6]

    Bobrowski, Functional analysis for probability and stochastic processes, Cambridge University Press, Cambridge, 2005

    A. Bobrowski, Functional analysis for probability and stochastic processes, Cambridge University Press, Cambridge, 2005

    work page 2005

  7. [7]

    Bui-Thanh, K

    T. Bui-Thanh, K. Willcox, and O. Ghattas, Model reduction for large-scale systems with high- dimensional parametric input space, SIAM Journal of Scientific Computing30 (2008), 3270–3288, doi:10.1137/070694855

    doi:10.1137/070694855 2008

  8. [9]

    Chinesta, A

    F. Chinesta, A. Huerta, G. Rozza, and K. Willcox, Model reduction methods, Encyclopaedia of Computational Mechanics (E. Stein, R. de Borst, and T. J. R. Hughes, eds.), vol. 1, John Wiley & Sons, Chichester, 2nd ed., 2017, Part 1. Fundamentals. Encyclopaedia of Computational Mechanics, doi:10.1002/9781119176817.ecm2110

    doi:10.1002/9781119176817.ecm2110 2017

Show all 55 references
  1. [10]

    Cohen and R

    A. Cohen and R. DeVore,Approximation of high-dimensional parametric pdes, Acta Numerica 24 (2015), 1–159, doi:10.1017/S0962492915000033

    2015 doi

  2. [11]

    Cohen, O

    N. Cohen, O. Sharir, Y. Levine, R. Tamari, D. Yakira, and A. Shashua,Analysis and design of con- volutional networks via hierarchical tensor decompositions, arXiv:1705.02302 [cs.LG], 2018, Available from: http://arxiv.org/abs/1705.02302, arXiv:1705.02302

    2018 arXiv

  3. [12]

    Cohen, O

    N. Cohen, O. Sharir, and A. Shashua, On the expressive power of deep learning: A tensor analysis, arXiv: 1509.05009 [cs.NE], 2016, Available from: http://arxiv.org/abs/1509.05009, arXiv:1509.05009

    2016 arXiv

  4. [14]

    Dautray and J.-L

    R. Dautray and J.-L. Lions, Spectral theory and applications, Mathematical Analysis and Nu- merical Methods for Science and Technology, vol. 3, Springer, Berlin, 1990, doi:10.1007/ 978-3-642-61529-0

    1990

  5. [15]

    Drohmann, B

    M. Drohmann, B. Haasdonk, and M. Ohlberger,Reduced basis approximation for non-linear para- metrized evolution equations based on empirical operator interpolation, SIAM Journal of Scientific Computing 34 (2012), A937–A969, doi:10.1137/10081157X

    2012 doi

  6. [16]

    Evensen,Data assimilation: The ensemble kalman filter, Springer, Berlin, 2007

    G. Evensen,Data assimilation: The ensemble kalman filter, Springer, Berlin, 2007

    2007

  7. [17]

    Falcó and A

    A. Falcó and A. Nouy,A proper generalized decomposition for the solution of elliptic problems in abstract form by using a functional Eckart-Young approach, J. Math. Anal. Appl.376 (2011), 469– 480, doi:10.1016/j.jmaa.2010.12.003

    2011 doi

  8. [18]

    Fleuret,The little book of deep learning, v1.1.1 ed., Université de Genève, 2023, Available from: https://fleuret.org/public/lbdl.pdf

    F. Fleuret,The little book of deep learning, v1.1.1 ed., Université de Genève, 2023, Available from: https://fleuret.org/public/lbdl.pdf

    2023

  9. [19]

    M. S. Floater, C. Manni, E. Sande, and H. Speleers,Best low-rank approximations and Kolmogorov n-widths, 2021, Available from:https://arxiv.org/abs/2007.13196, arXiv:2007.13196

    2021 arXiv

  10. [20]

    Fresca, L

    S. Fresca, L. Dede, and A. Manzoni,A comprehensive deep learning-based approach to reduced order modeling of nonlinear time-dependent parametrized pdes, 2020, Available from: https://arxiv. org/abs/2004.04001, arXiv:2004.04001

    2020 arXiv

  11. [21]

    Ghanem, D

    R. Ghanem, D. Higdon, and H. Owhadi (eds.),Handbook of uncertainty quantification, Springer, Berlin, 2017, doi:10.1007/978-3-319-12385-1

    2017 doi

  12. [22]

    Giraldi and A

    L. Giraldi and A. Nouy,Weakly intrusive low-rank approximation method for nonlinear parameter- dependent equations, SIAM Journal of Scientific Computing41(2019), A1777–A1792,doi:10.1137/ 17M1136924

    2019

  13. [23]

    Hackbusch,Tensor spaces and numerical tensor calculus, Springer, Berlin, 2012,doi:10.1007/ 978-3-642-28027-6

    W. Hackbusch,Tensor spaces and numerical tensor calculus, Springer, Berlin, 2012,doi:10.1007/ 978-3-642-28027-6

    2012

  14. [24]

    J. S. Hesthaven, G. Rozza, and B. Stamm,Certified reduced basis methods for parametrized partial differential equations, Springer, Berlin, 2016,doi:10.1007/978-3-319-22470-1

    2016 doi

  15. [25]

    E. T. Jaynes,Probability theory, the logic of science, Cambridge University Press, Cambridge, 2003

    2003

  16. [26]

    M. C. Kennedy and A. O’Hagan,Bayesian calibration of computer models, J. Royal Statist., Series B (2001), no. 63(3), 425–464

    2001

  17. [27]

    M. A. Kowalski, K. A. Sikorski, and F. Stenger,Selected Topics in Approximation and Computation, Oxford University Press, 1995

    1995

  18. [28]

    G.Kutyniok, P.Petersen, M.Raslan, andR.Schneider, A theoretical analysis of deep neural networks and parametric PDEs, 2019, Available from: https://arxiv.org/abs/1904.00377, arXiv:1904. 00377. 24 H.G. MATTHIES

    2019 arXiv

  19. [29]

    Latz,Bayesian inverse problems are usually well-posed, SIAM Review65 (2023), no

    J. Latz,Bayesian inverse problems are usually well-posed, SIAM Review65 (2023), no. 3, 831–865, doi:10.1137/23M1556435

    2023 doi

  20. [30]

    Masson-Delmotte, P

    V. Masson-Delmotte, P. Zhai, A. Pirani, S. L. Connors, C. Péan, S. Berger, N. Caud, Y. Chen, L. Goldfarb, M. I. Gomis, M. Huang, K. Leitzell, E. Lonnoy, J. B. R. Matthews, T. K. Maycock, T. Waterfield, O. Yelekçi, R. Yu, and B. Zhou (eds.),Ipcc, 2021: Climate change 2021: The ...

    2021 doi

  21. [31]

    H. G. Matthies,Uncertainty quantification with stochastic finite elements, Encyclopaedia of Com- putational Mechanics (E. Stein, R. de Borst, and T. J. R. Hughes, eds.), vol. 1, John Wiley & Sons, Chichester, 2007, Part 1. Fundamentals. Encyclopaedia of Computational Mechanics...

    2007 doi

  22. [32]

    Stein, R

    , Uncertainty quantification and Bayesian inversion, Encyclopaedia of Computational Mech- anics (E. Stein, R. de Borst, and T. J. R. Hughes, eds.), vol. 1, John Wiley & Sons, Chichester, 2nd ed., 2017, Part 1. Fundamentals. Encyclopaedia of Computational Mechanics,doi:10.1002/...

    2017

  23. [33]

    H. G. Matthies, Analysis of probabilistic and parametric reduced order models, Quantification of Uncertainty: Improving Efficiency and Technology (M. D’Elia, M. Gunzburger, and G. Rozza, eds.), Lecture Notes in Computational Science and Engineering, vol. 137, Springer, 2020, p...

    2020 doi

  24. [34]

    H. G. Matthies, A. Litvinenko, B. V. Rosić, and E. Zander,Bayesian parameter estimation via filtering and functional approximations, arXiv:1611.09293 [math.NA], 2016, Available from:http: //arxiv.org/abs/1611.09293, arXiv:1611.09293

    2016 arXiv

  25. [35]

    H. G. Matthies and R. Ohayon, Analysis of parametric models: Linear methods and ap- proximations, Advances in Computational Mathematics 45 (2019), 2555–2586, doi:10.1007/ s10444-019-09735-4

    2019

  26. [36]

    , Parametric models analysed with linear maps, Adv. Model. and Simul. in Eng. Sci.7 (2020), 41, doi:10.1186/s40323-020-00172-3

    2020 doi

  27. [37]

    H. G. Matthies, E. Zander, B. V. Rosić, and A. Litvinenko,Parameter estimation via conditional expectation: a Bayesian inversion., Advanced Modeling and Simulation in Engineering Sciences3 (2016), 24, doi:10.1186/s40323-016-0075-7

    2016 doi

  28. [38]

    Ibrahimbegović, ed.), Computational Methods in Applied Sciences, vol

    H.G.Matthies, E.Zander, B.V.Rosić, A.Litvinenko, andO.Pajonk, Inverse problems in a Bayesian setting, Computational Methods for Solids and Fluids — Multiscale Analysis, Probability Aspects and Model Reduction (A. Ibrahimbegović, ed.), Computational Methods in Applied Sciences,...

    2016 doi

  29. [39]

    K. P. Murphy,Machine learning: A probabilistic perspective, The MIT Press, Cambridge, MA, 2012

    2012

  30. [40]

    R. M. Neal,Bayesian learning for neural networks, LNS, vol. 118, Springer, 1995,doi:10.1007/ 978-1-4612-0745-0

    1995

  31. [41]

    N. H. Nelsen and A. M. Stuart,Operator learning using random features: a tool for scientific com- puting, 2024, Available from:https://arxiv.org/abs/2408.06526, arXiv:2408.06526

    2024 arXiv

  32. [42]

    O’Hagan and M

    A. O’Hagan and M. West (eds.),The Oxford Handbook of Applied Bayesian Analysis, Oxford Uni- versity Press, 2010

    2010

  33. [43]

    Pajonk, B

    O. Pajonk, B. V. Rosić, A. Litvinenko, and H. G. Matthies,A deterministic filter for non-Gaussian Bayesian estimation — applications to dynamical system estimation with noisy measurements, Phys- ica D 241 (2012), 775–788, doi:10.1016/j.physd.2012.01.001

    2012 doi

  34. [44]

    Pollard,A user’s guide to measure theoretic probability, Cambridge University Press, 2001

    D. Pollard,A user’s guide to measure theoretic probability, Cambridge University Press, 2001

    2001

  35. [45]

    Quarteroni, A

    A. Quarteroni, A. Manzoni, and F. Negri,Reduced basis methods for partial differential equations: An introduction,, Springer, Berlin, 2015,doi:10.1007/978-3-319-15431-2

    2015 doi

  36. [46]

    Quarteroni and G

    A. Quarteroni and G. Rozza (eds.), Reduced order methods for modeling and computational re- duction, MS&A — Modeling, Simulation & Applications, vol. 9, Berlin, Springer, 2014, doi: 10.1007/978-3-319-02090-7_8

    2014 doi

  37. [47]

    Raissi, P

    M. Raissi, P. Perdikaris, and G. E. Karniadakis,Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations, Journal of Computational Physics378 (2019), 686–707, doi:10.1016/j.jcp.2018.10.045

    2019 doi

  38. [48]

    B. V. Rosić, A. Kučerová, J. Sýkora, O. Pajonk, A. Litvinenko, and H. G. Matthies,Parameter identification in a probabilistic setting, Engineering Structures50 (2013), 179–196, doi:10.1016/j. engstruct.2012.12.029

    2013 doi

  39. [49]

    B. V. Rosić, A. Litvinenko, O. Pajonk, and H. G. Matthies,Sampling-free linear Bayesian update of polynomial chaos representations, Journal of Computational Physics231 (2012), 5761–5787, doi: 10.1016/j.jcp.2012.04.044. ANALYSING PARAMETRIC ROMS AND CONDITIONAL EXPECTATION 25

    2012 doi

  40. [50]

    Schmidhuber, Deep learning in neural networks: An overview, 2014, Available from: https: //arxiv.org/abs/1404.7828, arXiv:1404.7828

    J. Schmidhuber, Deep learning in neural networks: An overview, 2014, Available from: https: //arxiv.org/abs/1404.7828, arXiv:1404.7828

    2014 arXiv

  41. [51]

    Schwab and J

    C. Schwab and J. Zech, Deep learning in high dimension: Neural network expression rates for generalized polynomial chaos expansions in UQ, Anal. Appl. 17 (2019), 19–55, doi:10.1142/ S0219530518500203

    2019

  42. [52]

    S. K. Shivanand, B. Rosić, and H. G. Matthies,Stochastic modelling of symmetric positive definite material tensors, Journal of Computational Physics505 (2024), 112883,doi:10.1016/j.jcp.2024. 112883

    2024 doi

  43. [53]

    Soize and R

    C. Soize and R. Ghanem, Physics-constrained non-Gaussian probabilistic learning on manifolds, International Journal for Numerical Methods in Engineering121 (2019), 110–145, doi:10.1002/ nme.6202

    2019

  44. [54]

    Strang,Linear Algebra and Learning from Data, Wellesley - Cambridge Press, 2019

    G. Strang,Linear Algebra and Learning from Data, Wellesley - Cambridge Press, 2019

    2019

  45. [55]

    Vondřejc and H

    J. Vondřejc and H. G. Matthies,Accurate computation of conditional expectation for highly non- linear problems, SIAM/ASA Journal on Uncertainty Quantification7 (2019), 1349–1368, doi:10. 1137/18M1196674

    2019

  46. [56]

    G. Yang, Tensor programs I: Wide feedforward or recurrent neural networks of any architecture are gaussian processes, arXiv:1910.12478v3 [cs.NE], 2021, Available from:http://arxiv.org/abs/ 1910.12478, arXiv:1910.12478

    1910 arXiv

  47. [57]

    diagonal

    G. Yang and H. Salman,A fine-grained spectral perspective on neural networks, arXiv:1907.10599v1 [cs.LG], 2020, Available from:http://arxiv.org/abs/1907.10599, arXiv:1907.10599. Appendix A. Spectral Decomposition First a reminder of the finite dimensional theory: LetA =AT be a...

    1907 arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.