Pith. sign in

REVIEW 3 major objections 5 minor 47 references

Neural Spline Operators for Risk Quantification in Stochastic Systems

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read Neural Spline Operators learn direct maps from dynamics functions to safety probabilities, with a universal approximation guarantee.

desk verdict A genuinely new architecture with decent empirical results, but the universal approximation theorem as written is not proven. read the letter →

arxiv 2508.20288 v1 pith:MFNUJKS3 submitted 2025-08-27 eess.SY cs.LGcs.SY

classification eess.SYcs.LGcs.SY
keywords physics-informedneuraloperatorsriskquantificationsafetyprobabilityB-splinesstochasticdynamicalsystemsuniversalapproximationoperatorlearningmulti-agent
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 introduces Neural Spline Operators (NeSO), which it positions as the first physics-informed operator learning framework for risk quantification. The central claim is that one trained operator can take the system's dynamics function $f$ and safe-set parameters $\alpha$ as input and output the long-term safety probability $F(x,t)$ for any state and time horizon. This matters because sampling-based and PDE-based estimators become expensive or fragile when dynamics change in functional form, while existing physics-informed networks only accept fixed finite-dimensional parameter vectors. The paper proves a universal approximation theorem for NeSO and demonstrates the method on a system with randomly varying functional dynamics and on a 14-dimensional seven-agent system, reporting millisecond-scale online evaluation.

What carries the argument

The load-bearing object is the B-spline basis paired with learned control points. The safety probability is represented as $\hat F(x,t)=C\cdot B_{\ell,d}(x,t)$, where $C$ is a control-point tensor produced by a coefficient neural functional $G_\theta(f,\alpha)$ and $B_{\ell,d}$ is a tensor product of univariate B-spline bases in each state coordinate and in time. Two structural facts do the work: B-splines have closed-form derivatives, so the convection-diffusion residual can be formed without automatic differentiation, and the boundary control points of the tensor directly encode the initial condition and Dirichlet boundary conditions, so those constraints are enforced exactly rather than learned. The proof combines universal approximation of the neural operator layer with the fact that orthogonal projection onto the B-spline subspace is a bounded linear map that a feedforward network can reproduce.

What would settle it

Run a numerical experiment along a family of smooth dynamics functions that converge uniformly to a limit with a sharp drift change; compute the safety probability for each member with a high-accuracy PDE or sampling solver and check whether the probabilities converge. If they do not converge despite uniform convergence of the dynamics, the continuity premise behind the theorem fails, and NeSO would lose its guarantee on that family.

Watch

Extended reading notes

Core claim

On its own terms, the paper's discovery is that the risk-quantification problem can be reorganized as an operator learning problem: instead of re-solving the convection-diffusion equation for each dynamics function, a network learns the mapping $(f,\alpha)\mapsto F$. The network output is a tensor of B-spline control points; multiplying this tensor by the B-spline basis functions produces the predicted safety probability surface. B-spline derivatives are available in closed form, so the physics-based residual is cheap to evaluate, and the endpoint control points can be set by hand to satisfy initial and Dirichlet boundary conditions exactly. Theorem 1 asserts that for any accuracy $\varepsilon>0$ there exist B-spline counts and orders and network parameters $\theta$ such that $|F(x,t)-G_\theta(f,\alpha)(x,t)|\le\varepsilon$ uniformly over $f$, $\alpha$, $x$, and $t$, assuming the map from dynamics and safe sets to safety probabilities is continuous.

Load-bearing premise

The proof assumes that small changes in the dynamics function and safe-set parameters produce small changes in the safety probability; if that map has a jump, the universal approximation guarantee no longer applies.

Editorial extensions

If this is right

  • A single trained NeSO can evaluate safety probabilities for new, unseen dynamics functions in milliseconds, replacing repeated sampling or PDE solves with one forward pass.
  • Because it accepts a function as input, NeSO covers dynamics families whose functional form changes, rather than only parameterized variants of a fixed form.
  • Assigning initial and boundary conditions through B-spline control points removes the need for additional penalty terms that enforce those constraints during training.
  • The universal approximation theorem transfers to other PDE-constrained operator learning problems whose solution depends continuously on an input function.
  • In the multi-agent case, factorizing the full safety probability into sub-system probabilities lets NeSO reconstruct a 14-dimensional safety landscape from low-dimensional solves.

Reading between the lines

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

  • A stress test would push NeSO toward dynamics families that approach a discontinuous limit; the error behavior there would reveal whether the continuity assumption, not the architecture, is the operative bottleneck.
  • If the B-spline control-point trick is as generic as the proof suggests, the same construction could be applied to other PDE-constrained operators, fixing boundary conditions in the representation rather than through penalty terms.
  • The product-form reconstruction in the multi-agent study inherits an exact modal decoupling assumption; for non-symmetric or time-varying interaction graphs that factorization breaks, and NeSO would need a different decomposition to keep working.
  • The reported speed comparisons include training time for NeSO; a fair deployment comparison would amortize training across the number of query systems, which the paper's own cost formula starts to do.
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

3 major / 5 minor

Summary. The paper proposes Neural Spline Operators (NeSO), a physics-informed neural operator framework for estimating long-term safety probabilities of stochastic systems with varying functional dynamics. The safety probability is characterized as the solution of a convection-diffusion PDE, and NeSO maps the dynamics function f and safe-set parameter α to B-spline control points through a coefficient neural functional; the control points are multiplied by B-spline bases to produce the safety probability field. A universal approximation theorem (Theorem 1, Eq. (19)) is claimed. Two case studies are presented: a one-dimensional recovery-probability problem with randomly parameterized sinusoidal dynamics, and a 14-dimensional multi-agent mass-spring-damper system whose safety probability is reconstructed as a product of per-mode probabilities. The reported results show that NeSO improves accuracy and training time over FNO in the first case and achieves substantial online speedups over Monte Carlo and PDE solvers in the second.

Significance. If the universal approximation theorem were established as stated, the paper would be a meaningful step: it would be the first physics-informed operator learning method for risk quantification, with a B-spline output representation that enforces initial and boundary conditions and enables analytical derivatives. The experimental comparison is useful, the code is provided, and the online speedup claim is concrete. However, the theorem as written is not proven: the proof contains a category error in the approximation of the coefficient map, an unjustified uniformity over noncompact α∈R^ξ, and a false statement about orthogonality of B-splines. These are load-bearing because they support Eq. (19), the paper's central theoretical guarantee. The experimental claims are not deeply affected by these proof gaps, but the advertised guarantee currently exceeds what the proof establishes.

major comments (3)
  1. [Appendix A, proof of Theorem 1, Eq. (61)] The proof invokes the MLP universal approximation theorem [47] to assert that the coefficient map C*(·): Y→R^N can be approximated by a neural network, but [47] applies to continuous functions on finite-dimensional Euclidean spaces. Here Y=H^k(Ω2;R) is infinite-dimensional, and the proposed coefficient neural functional receives h=(f,α), not the function gh=W(h) on which C* is evaluated. No argument shows that the composition h↦C*(W(h)) is approximable by a neural operator layer followed by a linear layer. Since this step is used to bound the term ∥C*(gh)-Ĉ*(ĝh)∥ in Eq. (62), the proof does not establish the advertised universal approximation in Eq. (19).
  2. [Section V, Theorem 1 and Appendix A, Eq. (41)] The theorem quantifies uniformly over all α∈R^ξ and over f in a class that is not made precise, whereas the FNO universal approximation theorem [43] used in Eq. (41) requires the input set to be compact and the operator W to be continuous on that set. The manuscript only assumes that K1 and K2 are subsets of Sobolev spaces and states in footnote 2 that continuity 'holds with regularity conditions,' without stating those conditions. To make Eq. (19) follow from [43], the statement must restrict (f,α) to a compact product set K1×K2 and state the regularity conditions under which (f,α)↦F is continuous.
  3. [Appendix A, footnote 4 and Eqs. (42)-(47)] The footnote claims that B-spline basis functions are orthogonal to each other; this is false. B-spline bases are locally supported and linearly independent but not orthogonal in L2, so the asserted invertibility of B_φ 'since φ_j are orthogonal' is not justified. The argument can be repaired by using the positive definiteness of the Gram matrix of a B-spline basis, but as written the proof contains a false mathematical claim at a point used to define the optimal coefficients C*(gh).
minor comments (5)
  1. [Section I] The word 'inital' in the introduction should be corrected to 'initial.'
  2. [Section V, Theorem 1] The statement 'for any function f∈F' uses an undefined symbol F; it should be f∈K1 or another explicitly defined class, and the notation should be made consistent with the earlier assumption f∈K1⊂C(Ω1;R^n).
  3. [Section IV, Eq. (14)] The PDE residual uses d/dx and d^2/dx^2 notation for a vector state x; this should be written with ∇_x and trace(σσ^T ∇_x^2 F) so that the residual is well-defined for n>1 as well as for the scalar case.
  4. [Section VI-A, Eqs. (21) and (25)] The dynamics function in Eq. (21) depends only on t, while the problem formulation in Eq. (1) allows x-dependence; the paper should state explicitly that this case study uses time-dependent dynamics, so that S(τ) in Eq. (25) is independent of the state trajectory.
  5. [Section VI-B, Table II] The reported NeSO online time of 56.47 s should be separated into the network evaluation time and any additional reconstruction overhead, and the 3999 s training time should be clearly labeled as offline, since the advertised speedup depends on this offline/online split.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the universal approximation claim is supported by external approximation theorems, not by fitting or by self-defined targets.

full rationale

The paper's central claim, Theorem 1, is an approximation guarantee for the NeSO architecture. Its proof in Appendix A reduces the problem to two external ingredients: the universal approximation theorem for neural operators (FNO/DeepONet, cited as [43] and [27]) and the classical approximation power of B-spline bases. The proof then uses a projection argument to convert a neural-operator approximation of the input-to-solution map into a spline-coefficient representation. This is a derivation from established results rather than a circular definition. The self-citations in the paper—[3] for the PDE characterization of safety probabilities, [5] for B-spline approximation, and [46] for the modal decomposition in the multi-agent case—are to published or externally checkable results whose assumptions do not include the theorem being proved. The multi-agent prediction is reconstructed from sub-system probabilities using the external factorization of [46, Theorem 1] and is validated against independent Monte Carlo simulation, not against the training target. No parameter is fitted to a subset of data and then renamed as a prediction. The proof does contain gaps that are correctness risks, notably the invocation of the finite-dimensional MLP universal approximation theorem [47] for the infinite-dimensional coefficient map C* : Y -> R^N, and the quantification over noncompact alpha in R^xi. However, those are mathematical-rigor issues, not circularity: they do not make the conclusion equivalent to the inputs by construction. Under the hard rules, proof gaps and nonstandard-consensus concerns are not circularity findings. Therefore an honest non-finding is appropriate, with score 0.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The central claim is an existence proof and an empirical method; the ledger is dominated by standard operator approximation theorems and domain-specific PDE facts pulled from prior literature. The main domain assumptions are continuity of the risk functional and exactness of the modal decomposition, both imported from external results. No new entities are introduced.

free parameters (4)
  • Loss weights w_p, w_d, w_ICBC = w_p=1, w_d=3, w_ICBC=10 in case study VI-A; not specified for VI-B
    Chosen by hand with a general heuristic in Section IV; they affect the balance between physics and data losses.
  • B-spline knot count and order = not reported in text
    The universal approximation theorem only guarantees existence; the experiments do not state the number of control points ℓ or B-spline orders d used.
  • FNO architecture hyperparameters = 8 Fourier modes, width 32, 3 spectral layers
    Specified for case study VI-A; these are model capacity choices that affect accuracy and training time.
  • Training configuration = Adam, lr=0.001, 500 epochs
    Standard choices; no sensitivity analysis is reported.
assumptions (6)
  • standard math Universal approximation theorems for neural operators (FNO [43], DeepONet [27]) and MLPs [47]
    Used in Appendix A to prove the coefficient neural functional can approximate the projection coefficients.
  • domain assumption Continuity of the map (f,α) -> F
    Assumed in Section V footnote 2; required for the universal approximation theorem.
  • domain assumption Safety probability satisfies convection-diffusion equation (4) with ICBC (5)
    Borrowed from [3]; the entire physics loss rests on this PDE.
  • domain assumption Modal decomposition F = Π_k F_k from [46, Theorem 1]
    Used in Section VI-B to reduce the 14D system to 7 independent 2D subsystems.
  • standard math B-splines are universal approximators
    Invoked in Appendix A, Eq. (67), with citation to the authors' own prior work [5]; the property itself is standard.
  • standard math B-spline Gram matrix is positive definite
    Needed for the projection argument; the paper incorrectly states orthogonality, but linear independence is enough for invertibility.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Neural Spline Operators for Risk Quantification in Stochastic Systems." pith.science (2026). https://pith.science/paper/MFNUJKS3

@misc{pith2026250820288,
  author       = {Pith},
  title        = {Pith review of: Neural Spline Operators for Risk Quantification in Stochastic Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MFNUJKS3}},
  note         = {Machine review of arXiv:2508.20288}
}
read the original abstract

Accurately quantifying long-term risk probabilities in diverse stochastic systems is essential for safety-critical control. However, existing sampling-based and partial differential equation (PDE)-based methods often struggle to handle complex varying dynamics. Physics-informed neural networks learn surrogate mappings for risk probabilities from varying system parameters of fixed and finite dimensions, yet can not account for functional variations in system dynamics. To address these challenges, we introduce physics-informed neural operator (PINO) methods to risk quantification problems, to learn mappings from varying \textit{functional} system dynamics to corresponding risk probabilities. Specifically, we propose Neural Spline Operators (NeSO), a PINO framework that leverages B-spline representations to improve training efficiency and achieve better initial and boundary condition enforcements, which are crucial for accurate risk quantification. We provide theoretical analysis demonstrating the universal approximation capability of NeSO. We also present two case studies, one with varying functional dynamics and another with high-dimensional multi-agent dynamics, to demonstrate the efficacy of NeSO and its significant online speed-up over existing methods. The proposed framework and the accompanying universal approximation theorem are expected to be beneficial for other control or PDE-related problems beyond risk quantification.

Figures

Figures reproduced from arXiv: 2508.20288 by the authors.

Figure 1
Figure 1. Diagram of the proposed neural spline operator (NeSO). The coefficient neural functional takes dynamics function [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Prediction results on a random testing function [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Safety probability estimation for the multi-agent [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 38 canonical work pages

  1. [47]

    Multilayer feedforward networks are universal approximators,

    K. Hornik, M. Stinchcombe, and H. White, “Multilayer feedforward networks are universal approximators,” Neural networks, vol. 2, no. 5, pp. 359–366, 1989. APPENDIX A. Proof of Theorem 1 Proof. (Theorem 1) We state the general proof of the universal approximation property of neural operators with basis functions, for learning continuous mappings between So...

  2. [43]

    On universal approximation and error bounds for fourier neural operators,

    N. Kovachki, S. Lanthaler, and S. Mishra, “On universal approximation and error bounds for fourier neural operators,” Journal of Machine Learning Research, vol. 22, no. 290, pp. 1–76, 2021

  3. [1]

    Myopi- cally verifiable probabilistic certificate for long-term safety,

    Z. Wang, H. Jing, C. Kurniawan, A. Chern, and Y . Nakahira, “Myopi- cally verifiable probabilistic certificate for long-term safety,” in 2022 American Control Conference (ACC) . IEEE, 2022, pp. 4894–4900

  4. [2]

    Rubino and B

    G. Rubino and B. Tuffin, Rare event simulation using Monte Carlo methods. John Wiley & Sons, 2009

  5. [3]

    Safe control in the presence of stochastic uncertainties,

    A. Chern, X. Wang, A. Iyer, and Y . Nakahira, “Safe control in the presence of stochastic uncertainties,” in 2021 60th IEEE Conference on Decision and Control (CDC) . IEEE, 2021, pp. 6640–6645

  6. [4]

    A generalizable physics-informed learning framework for risk probability estimation,

    Z. Wang and Y . Nakahira, “A generalizable physics-informed learning framework for risk probability estimation,” in Learning for Dynamics and Control Conference. PMLR, 2023, pp. 358–370

  7. [5]

    Physics- informed deep b-spline networks for dynamical systems,

    Z. Wang, R. Romagnoli, J. Ratchford, and Y . Nakahira, “Physics- informed deep b-spline networks for dynamical systems,” arXiv preprint arXiv:2503.16777, 2025

  8. [6]

    Neural operator: Learning maps be- tween function spaces with applications to pdes,

    N. Kovachki, Z. Li, B. Liu, K. Azizzadenesheli, K. Bhattacharya, A. Stuart, and A. Anandkumar, “Neural operator: Learning maps be- tween function spaces with applications to pdes,” Journal of Machine Learning Research, vol. 24, no. 89, pp. 1–97, 2023

Show all 47 references
  1. [7]

    Physics-informed neural operator for learning partial differential equations,

    Z. Li, H. Zheng, N. Kovachki, D. Jin, H. Chen, B. Liu, K. Aziz- zadenesheli, and A. Anandkumar, “Physics-informed neural operator for learning partial differential equations,” ACM/JMS Journal of Data Science, vol. 1, no. 3, pp. 1–27, 2024

  2. [8]

    Neural operators for accelerating scientific simulations and design,

    K. Azizzadenesheli, N. Kovachki, Z. Li, M. Liu-Schiaffini, J. Kossaifi, and A. Anandkumar, “Neural operators for accelerating scientific simulations and design,” Nature Reviews Physics , vol. 6, no. 5, pp. 320–328, 2024

  3. [9]

    J. H. Ahlberg, E. N. Nilson, and J. L. Walsh, The Theory of Splines and Their Applications: Mathematics in Science and Engineering: A Series of Monographs and Textbooks, Vol. 38. Elsevier, 2016, vol. 38

  4. [10]

    Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations,

    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 physics , vol. 378, pp. 686–707, 2019

  5. [11]

    Adaptive safe con- trol for driving in uncertain environments,

    S. Gangadhar, Z. Wang, H. Jing, and Y . Nakahira, “Adaptive safe con- trol for driving in uncertain environments,” in 2022 IEEE Intelligent Vehicles Symposium (IV). IEEE, 2022, pp. 1662–1668

  6. [12]

    3d dynamic walking on stepping stones with control barrier functions,

    Q. Nguyen, A. Hereid, J. W. Grizzle, A. D. Ames, and K. Sreenath, “3d dynamic walking on stepping stones with control barrier functions,” in 2016 IEEE 55th Conference on Decision and Control (CDC) . IEEE, 2016, pp. 827–834

  7. [13]

    Monte-carlo based uncertainty analysis: Sampling effi- ciency and sampling convergence,

    H. Janssen, “Monte-carlo based uncertainty analysis: Sampling effi- ciency and sampling convergence,” Reliability Engineering & System Safety, vol. 109, pp. 123–132, 2013

  8. [14]

    Sequential monte carlo for rare event estimation,

    F. C ´erou, P. Del Moral, T. Furon, and A. Guyader, “Sequential monte carlo for rare event estimation,” Statistics and computing , vol. 22, no. 3, pp. 795–808, 2012

  9. [15]

    Estimation of small failure probabilities in high dimensions by subset simulation,

    S.-K. Au and J. L. Beck, “Estimation of small failure probabilities in high dimensions by subset simulation,” Probabilistic engineering mechanics, vol. 16, no. 4, pp. 263–277, 2001

  10. [16]

    Multilevel monte carlo methods,

    M. B. Giles, “Multilevel monte carlo methods,” Acta numerica, vol. 24, pp. 259–328, 2015

  11. [17]

    Dhatt, E

    G. Dhatt, E. Lefranc ¸ois, and G. Touzot, Finite element method. John Wiley & Sons, 2012

  12. [18]

    Finite volume methods,

    R. Eymard, T. Gallou ¨et, and R. Herbin, “Finite volume methods,” Handbook of numerical analysis , vol. 7, pp. 713–1018, 2000

  13. [19]

    An assessment of solvers for algebraically stabilized discretizations of convection–diffusion– reaction equations,

    A. Jha, O. P ´artl, N. Ahmed, and D. Kuzmin, “An assessment of solvers for algebraically stabilized discretizations of convection–diffusion– reaction equations,” Journal of Numerical Mathematics, vol. 31, no. 2, pp. 79–103, 2023

  14. [20]

    Eliminating gibbs phenomena: A non-linear petrov–galerkin method for the convection– diffusion–reaction equation,

    P. Houston, S. Roggendorf, and K. G. van der Zee, “Eliminating gibbs phenomena: A non-linear petrov–galerkin method for the convection– diffusion–reaction equation,” Computers & Mathematics with Appli- cations, vol. 80, no. 5, pp. 851–873, 2020

  15. [21]

    Four- castnet: A global data-driven high-resolution weather model using adaptive fourier neural operators,

    J. Pathak, S. Subramanian, P. Harrington, S. Raja, A. Chattopadhyay, M. Mardani, T. Kurth, D. Hall, Z. Li, K. Azizzadenesheli et al., “Four- castnet: A global data-driven high-resolution weather model using adaptive fourier neural operators,” arXiv preprint arXiv:2202.11214 , 2022

  16. [22]

    Fourier neural operator approach to large eddy simulation of three-dimensional turbulence,

    Z. Li, W. Peng, Z. Yuan, and J. Wang, “Fourier neural operator approach to large eddy simulation of three-dimensional turbulence,” Theoretical and Applied Mechanics Letters , vol. 12, no. 6, p. 100389, 2022

  17. [23]

    Learning the solution operator of parametric partial differential equations with physics-informed deeponets,

    S. Wang, H. Wang, and P. Perdikaris, “Learning the solution operator of parametric partial differential equations with physics-informed deeponets,” Science advances, vol. 7, no. 40, p. eabi8605, 2021

  18. [24]

    Fourier neural operator for parametric partial differential equations,

    Z. Li, N. Kovachki, K. Azizzadenesheli, B. Liu, K. Bhattacharya, A. Stuart, and A. Anandkumar, “Fourier neural operator for parametric partial differential equations,” arXiv preprint arXiv:2010.08895, 2020

  19. [25]

    Neural operator: Graph kernel network for partial differential equations,

    ——, “Neural operator: Graph kernel network for partial differential equations,” arXiv preprint arXiv:2003.03485 , 2020

  20. [26]

    Spectral neural operators,

    V . S. Fanaskov and I. V . Oseledets, “Spectral neural operators,” in Doklady Mathematics , vol. 108, no. Suppl 2. Springer, 2023, pp. S226–S232

  21. [27]

    Deeponet: Learning non- linear operators for identifying differential equations based on the universal approximation theorem of operators,

    L. Lu, P. Jin, and G. E. Karniadakis, “Deeponet: Learning non- linear operators for identifying differential equations based on the universal approximation theorem of operators,” arXiv preprint arXiv:1910.03193, 2019

  22. [28]

    Improving physics-informed deeponets with hard constraints,

    R. Brecht, D. R. Popovych, A. Bihlo, and R. O. Popovych, “Improving physics-informed deeponets with hard constraints,” arXiv preprint arXiv:2309.07899, 2023

  23. [29]

    Reproducing kernel triangular b-spline-based fem for solving pdes,

    Y . Jia, Y . Zhang, G. Xu, X. Zhuang, and T. Rabczuk, “Reproducing kernel triangular b-spline-based fem for solving pdes,” Computer Methods in Applied Mechanics and Engineering , vol. 267, pp. 342– 358, 2013

  24. [30]

    Variational formulation based on duality to solve partial differential equations: Use of b-splines and machine learning approximants,

    N. Sukumar and A. Acharya, “Variational formulation based on duality to solve partial differential equations: Use of b-splines and machine learning approximants,” arXiv preprint arXiv:2412.01232 , 2024

  25. [31]

    Functional networks for b- spline surface reconstruction,

    A. Iglesias, G. Echevarr ´ıa, and A. G ´alvez, “Functional networks for b- spline surface reconstruction,” Future Generation Computer Systems , vol. 20, no. 8, pp. 1337–1353, 2004

  26. [32]

    Modeling nonlinear systems using the tensor network b-spline and the multi-innovation identification the- ory,

    Y . Wang, S. Tang, and M. Deng, “Modeling nonlinear systems using the tensor network b-spline and the multi-innovation identification the- ory,” International Journal of Robust and Nonlinear Control , vol. 32, no. 13, pp. 7304–7318, 2022

  27. [33]

    Dif- ferentiable spline approximations,

    M. Cho, A. Balu, A. Joshi, A. Deva Prasad, B. Khara, S. Sarkar, B. Ganapathysubramanian, A. Krishnamurthy, and C. Hegde, “Dif- ferentiable spline approximations,” Advances in neural information processing systems, vol. 34, pp. 20 270–20 282, 2021

  28. [34]

    Learning feedforward control using a dilated b-spline network: Frequency domain analysis and design,

    Y . Chen, K. L. Moore, and V . Bahl, “Learning feedforward control using a dilated b-spline network: Frequency domain analysis and design,” IEEE Transactions on neural networks , vol. 15, no. 2, pp. 355–366, 2004

  29. [35]

    Building hybrid b- spline and neural network operators,

    R. Romagnoli, J. Ratchford, and M. H. Klein, “Building hybrid b- spline and neural network operators,” in 2024 IEEE 63rd Conference on Decision and Control (CDC) . IEEE, 2024, pp. 3611–3617

  30. [36]

    Deep neural networks for smooth approximation of physics with higher order and continuity b-spline base functions,

    K. Doległo, A. Paszy ´nska, M. Paszy ´nski, and L. Demkowicz, “Deep neural networks for smooth approximation of physics with higher order and continuity b-spline base functions,” arXiv preprint arXiv:2201.00904, 2022

  31. [37]

    A best-fitting b-spline neural network approach to the prediction of advection– diffusion physical fields with absorption and source terms,

    X. Zhu, J. Liu, X. Ao, S. He, L. Tao, and F. Gao, “A best-fitting b-spline neural network approach to the prediction of advection– diffusion physical fields with absorption and source terms,” Entropy, vol. 26, no. 7, p. 577, 2024

  32. [38]

    Multipole graph neural operator for parametric partial differential equations,

    Z. Li, N. Kovachki, K. Azizzadenesheli, B. Liu, A. Stuart, K. Bhat- tacharya, and A. Anandkumar, “Multipole graph neural operator for parametric partial differential equations,” Advances in Neural Infor- mation Processing Systems , vol. 33, pp. 6755–6766, 2020

  33. [39]

    U-no: U-shaped neural operators,

    M. A. Rahman, Z. E. Ross, and K. Azizzadenesheli, “U-no: U-shaped neural operators,” arXiv preprint arXiv:2204.11127 , 2022

  34. [40]

    Seismic wave propagation and inversion with neural operators,

    Y . Yang, A. F. Gao, J. C. Castellanos, Z. E. Ross, K. Azizzadenesheli, and R. W. Clayton, “Seismic wave propagation and inversion with neural operators,” The Seismic Record , vol. 1, no. 3, pp. 126–134, 2021

  35. [41]

    Broadband ground-motion synthesis via generative adver- sarial neural operators: Development and validation,

    Y . Shi, G. Lavrentiadis, D. Asimaki, Z. E. Ross, and K. Azizzade- nesheli, “Broadband ground-motion synthesis via generative adver- sarial neural operators: Development and validation,” Bulletin of the Seismological Society of America , vol. 114, no. 4, pp. 2151–2171, 2024

  36. [42]

    The computation of all the derivatives of a b-spline basis,

    K. R. Butterfield, “The computation of all the derivatives of a b-spline basis,” IMA Journal of Applied Mathematics, vol. 17, no. 1, pp. 15–25, 1976

  37. [44]

    L. C. Evans, Partial differential equations. American mathematical society, 2022, vol. 19

  38. [45]

    N. V . Krylov, Lectures on elliptic and parabolic equations in H ¨older spaces. American Mathematical Soc., 1996, vol. 12

  39. [46]

    Orthogonal modal representa- tion in long-term risk quantification for dynamic multi-agent systems,

    R. Yasunaga, Y . Nakahira, and Y . Hori, “Orthogonal modal representa- tion in long-term risk quantification for dynamic multi-agent systems,” IEEE Control Systems Letters , 2024

Pith tools

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