betaFeedback

WorksMathematics

The Segal-Bargmann Transform in Neural Partial Differential Equations

Abstract

We introduce a novel neural operator method based on Segal-Bargmann transform and Bargmann-Fock space for learning the solution operators of partial differential equations (PDEs). The method brings data encoding, latent-space representation, operator learning, and decoding into an unifed framework. From only finitely many observations, we establish a sup-norm reconstruction error of order L−12 L^{-\frac{1}{2}} , where L is the number of complex sampling points that are decoupled with the original samplings and observations. The reconstruction error is independent of the spatial dimension and the number of original sampling points. We show that this new mesh-free Segal-Bargmann Neural Operator approach has superior performance compared to existing NO approaches like FNO, BelNet, GNOT, etc.,\ particularly in terms of pointwise control sup⁡x∣u^NO−utruth∣ \sup_{x}\lvert\hat{u}_{NO} - u_{truth}\rvert for solutions of PDEs Lu=f \mathcal{L}u = f which provides pointwise accuracy and interpretability as classical numerical solvers. We also demonstrates the reconstructions of the solution operators for higher dimensional nonlinear PDEs, which suggest that our proposed method has great potential to be applied to more challenging PDEs and to even higher spatial dimensions.

Introduction

Let Ω\Omega be an open subset of Rn\mathbb{R}^n. Consider the partial differential equation Lu=f\mathcal{L}u=f in Ω⊂Rn\Omega\subset\mathbb{R}^n where u,f∈L2(Rn;R)u,f \in L^2(\mathbb{R}^n;\mathbb{R}) and L\mathcal{L} is some smooth differential operator, possibly with boundary and initial conditions. In this article, we are interesting the following neural operator learning question to study L−1\mathcal{L}^{-1}:

Question: Given only a finite collection of observations {u(k)(xi),f(k)(xi)}k=1K\{u^{(k)}(x_i),f^{(k)}(x_i)\}_{k=1}^{K} of unknown ground truth pairing of functions {u(k),f(k)}k=1K\{u^{(k)},f^{(k)}\}_{k=1}^{K} on finitely many grids {xi}i=1N⊂Rn\{x_i\}_{i=1}^{N}\subset\mathbb{R}^n, where each u(k)(x)u^{(k)}(x) solves Lu(k)(x)=f(k)(x)\mathcal{L}u^{(k)}(x)=f^{(k)}(x) with prescribed boundary conditions. Can we learn a neural operator GNOG_{NO} that takes the sampled values f=(f(x1),…,f(xN))\mathbf{f}=(f(x_1),\ldots,f(x_N)) of a new input function ff from the prescribed function class, and reconstruct u^(x):=GNO(f)(x)\hat{u}(x):=G_{NO}(\mathbf{f})(x) that approximates the corresponding solution uu of Lu=f\mathcal{L}u=f at any query locations x∈Ωx\in\Omega?

The classical numerical solvers are designed to address this type of problem, but their computational cost can grow rapidly with spatial dimension; in addition, a new input instance generally requires a separate numerical solve of the underlying PDE. Over the last several decades, neural operator methods with deep learning architecture have been developed to learn the solution map for an entire family of PDE problems [36, 29]. A summary of representative approaches is given in §2.

However, neural operator methods have several major issues: generalization remains a major limitation of neural operators, especially for out-of-distribution inputs, unseen discretizations, or different sampling grids; the curse of dimensionality is still not overcome; existing methods also lack the same level of stability guarantees, rigorous error control, and interpretability as classical numerical solvers. Moreover, a trained neural operator is usually tied to a specific PDE family and parameter regime, so substantial retraining may be required when the underlying equation or physical setting changes. For example, many numerical comparisons only report relative L2L^2 errors, which alone do not control the error uniformly in space; the standard FFT-based implementation FNO [32] require sampling over uniform grids, other methods require associated preferred patterns of grids; a full tensor-product Fourier representation requires exponentially many modes as the spatial dimension increases, although with a fixed cutoff in each coordinate; in the standard DeepONet architecture, the latent basis is learned implicitly by the trunk network rather than prescribed from an analytically interpretable function space, which can make the resulting representation difficult to interpret.

In this article, we introduce a novel neural operator method based on the Segal-Bargmann transform, where the learning process is carried out in a new space with good properties. In addition, we also show that the combination of the Segal-Bargmann transform and the RKHS approximation serves as a good encoder and a finite dimensional representation of the original discrete data that are coming from evaluation representations of functions. In addition to the classical L2L^2-type convergence rate, we give an L∞L^\infty-type convergence rate.

The Segal-Bargmann transform was initially introduced by the independent work of Irving Segal [42] and Valentine Bargmann [2] in the early 1960s, where Segal gave the general framework on Hilbert spaces, and Bargmann published the explicit integral formula in 1961 [20]. It maps square-integrable functions on Rn\mathbb{R}^n unitarily onto the Bargmann-Fock space of entire functions on Cn\mathbb{C}^n with a Gaussian-weighted norm [2]. Its principal applications include coherent-state representations and quantization in quantum mechanics for detailed analysis of annihilation and creation operators acting on a many-particle quantum state [20], as well as time-frequency analysis through its relation to the Gaussian-windowed short-time Fourier transform [18]. Phase space is natural here because, up to scaling and sign conventions, the real and imaginary coordinates label the position and momentum (or location and frequency) of Gaussian wave packets, allowing the transform to describe where oscillatory features occur as well as their frequencies [20, 18].

We denote by Fα2(Cn)\mathcal{F}^{2}_{\alpha}(\mathbb{C}^n) the Bargmann-Fock space. We also denote by Bα:L2(Rn;C)→Fα2(Cn)\mathcal{B}_{\alpha}: L^{2}(\mathbb{R}^{n};\mathbb{C}) \to\mathcal{F}^{2}_{\alpha}(\mathbb{C}^{n}) and Bα−1:Fα2(Cn)→L2(Rn;C)\mathcal{B}_{\alpha}^{-1}: \mathcal{F}^{2}_{\alpha}(\mathbb{C}^{n}) \to L^{2}(\mathbb{R}^{n};\mathbb{C}) the Segal-Bargmann transform and inverse-Segal-Bargmann transform, respectively. The precise definitions will be given in § 3. There is no difference if we restrict the inputs to real-valued functions. In contrast with the current neural operator approaches, we would like to emphasize the advantages of our Segal-Bargmann neural operator approach:

(a) When f∈L2(Rn;R)∩C∞(Rn;R)f \in L^{2}(\mathbb{R}^{n};\mathbb{R}) \cap C^{\infty}(\mathbb{R}^{n};\mathbb{R}) and both Assumption A and Assumption B given in §5 hold, Theorem 5.4 says the pointwise difference between the estimate u^=GNO(f)\hat{u} = \mathcal{G}_{\mathrm{NO}}(f) and the ground truth uu satisfies the dimension-free convergence rate

E[sup⁡x∈Rn∣u^−u∣]=O(L−12)\mathbb{E}\left[\sup_{x\in\mathbb{R}^{n}}|\hat{u}-u|\right] = O\left(L^{-\frac{1}{2}}\right)

with standard assumptions of the universal approximation theorem of neural operators on Hilbert spaces. Here LL is the number of complex sampling points {zl}l=1L⊂Cn\{z_l\}_{l=1}^{L} \subset\mathbb{C}^{n}, and is independent of the number of original sampling points {xi}i=1N\{x_i\}_{i=1}^{N}. This canonical Monte-Carlo convergence rate is quite general in the sense that it is applicable to almost all partial differential equations, and has almost no requirement of the original grids. One of the advantages of this Segal-Bargmann NO approach is that we can select our complex grids as “probes” or “detectors” freely.

(b) Assumption A is too restrictive. If we drop the Assumption A and assume that f∈Hs(Rn;R)∩C∞(Rn;R)f \in H^{s}(\mathbb{R}^{n};\mathbb{R}) \cap C^{\infty}(\mathbb{R}^{n};\mathbb{R}). If s>n2s > \frac{n}{2}, then we give the dimension-free pointwise convergence rate

E[sup⁡x∈Rn∣u^−u∣]=O(L−12)\mathbb{E}\left[\sup_{x\in\mathbb{R}^{n}}|\hat{u}-u|\right] = O\left(L^{-\frac{1}{2}}\right)

in Theorem 5.5, provided that Assumption B holds. If s<n2s < \frac{n}{2}, then a natural consequence of Theorem 5.3 and universal approximation theorem of neural operators will imply that

E[∥u−u^∥L2(Rn;R)]=O(L−s2n).\mathbb{E}\left[\|u-\hat{u}\|_{L^{2}(\mathbb{R}^{n};\mathbb{R})}\right] = O\left(L^{-\frac{s}{2n}}\right).

We give a theorem in the Appendix, which demonstrates that our Assumption B can be achieved with some hypotheses and is therefore a reasonable assumption in general.

(c) The L−12L^{- \frac{1}{2}} reconstruction errors are dimension-free under the stated hypotheses, which indicates that our Segal-Bargmann neural operator approach partially overcomes the curse of dimensionality. In principle, we can approach to the true solution of the PDE by increasing the complex grids L→∞L \to\infty, with just fixed number of observations on {xi}i=1N\{x_i\}_{i=1}^{N}. We give numerical results for the highly nonlinear PDE −Δu+eu+u3=f-\Delta u + e^u + u^3 = f in Ω\Omega, u=0u = 0 on ∂Ω\partial\Omega for Ω⊂R5\Omega\subset\mathbb{R}^5. Based on our best understanding, the state-of-the-art NO approaches for unknown differential operator L\mathcal{L} is spatial dimension 3.

(d) The Segal-Bargmann neural operator approach integrates the information recovery and data encoding, the latent space representation, the learning process and the data decoding into a unified framework, summarized by the following diagram:

Figure 5.

(e) Our algorithm is fast and accurate since our learning process is undertaken in the Bargmann-Fock space, which is an RKHS consisting of complex analytic functions. The RKHS property and the rigidity of holomorphic functions allow us to use small neural network to achieve high accuracy with low computational cost. Our experiments report not only small L2L^2-type errors, but also much smaller EX[∣u^(X)−u(X)∣∣u(X)∣]\mathbb{E}_X\left[\frac{|\hat{u}(X)-u(X)|}{|u(X)|}\right] and sup⁡x∣u^(x)−u(x)∣∣u(x)∣\sup_x \frac{|\hat{u}(x)-u(x)|}{|u(x)|} errors than the NO baselines considered, i.e. we also measure the typical (expected) difference between the estimated solution u^\hat{u} and the ground truth uu. Our algorithm uses the Gaussian-windowed Fourier-transform nature of the Segal-Bargmann transform, which ensures the stability and high accuracy of our algorithm.

(f) Both the Segal-Bargmann and inverse-Segal-Bargmann transforms are unitary transformations. The key feature of this approach is that we decouple the observations of ff in the original space at {xi}∈Rn\{x_i\} \in\mathbb{R}^n from the learning process with encoding and representation in the new space via new grids {zl}⊂Cn\{z_l\} \subset\mathbb{C}^n. Precisely speaking, the Segal-Bargmann transform serves not only as a unitary information transformation, but also as an encoder, while the RKHS representation of the encoded data in Bargmann-Fock space is designed as the latent space representation of the original data. The inverse-Segal-Bargmann transform can be designed as an exact decoder at the continuous level.

This article is organized as follows: In §2, we give a comprehensive summary of the state-of-the-art data-driven PDE approaches, especially over the last decade. In §3, we give the necessary definitions of the Segal-Bargmann transform, the Bargmann-Fock space, the RKHS kernel, and an associated orthonormal basis in the Bargmann-Fock space that will be used in later sections. In §4, we give an example in §4.1 of how the Segal-Bargmann transform can convert a partial differential equation with solutions in L2(Rn;R)L^2(\mathbb{R}^n;\mathbb{R}) into a new, complex-valued partial differential equation in Fa2(Cn)\mathcal{F}_a^2(\mathbb{C}^n), where two solutions correspond through the Segal-Bargmann transform. We give the problem setting in § 4.2. In § 4.3, we give a two-step algorithm where we study the encoder WW first, and then study the neural operator in the Bargmann-Fock space. In § 4.4, we give an end-to-end algorithm that combines the study of the encoder and the neural operator together. In § 5, we give theoretical results following the problem setting in § 4. With Assumption AA and Assumption BB, we show in Theorem 5.4 that the error bound between the estimate u^\hat{u} and the solution uu is dimension-free in the sense of the sup-norm, which is at least at the level of classical Monte-Carlo rate O(L−1/2)O\left(L^{-1/2}\right) where LL is the number of the complex sampling points we choose, decoupled from the original real grids. Once we drop the Assumption AA and consider the most general input ff and solution uu, we show in Theorem 5.5 that the same conclusion holds by assuming that ff and uu belong to suitable Sobolev spaces. In § 6, we give numerical results for 2D Darcy flow, 1D Burgers’ equation with appropriate initial and boundary conditions. For both equations above, we also provide benchmark comparisons with established methods such as FNO, GNOT, Belnet, etc. for both uniform grids and scattered points. We also give numerical results for two nonlinear PDEs in spatial dimension 5.

Related Work in Neural PDEs

Our Segal-Bargmann neural operator approach is a neural operator that uniquely integrates the Segal-Bargmann encoding, operator learning in a finite-dimensional Bargmann-Fock RKHS space, and inverse Segal-Bargmann decoding into a unified framework for learning PDE solution operators. Over the past decade, learning-based PDE solvers have been significantly developed along two main directions: approximating the solution of an individual problem and learning the operator that maps coefficients, forcing terms, or initial and boundary conditions to solutions. These approaches differ in their representations, physical constraints, and use of numerical algorithms.

Physics-informed neural networks (PINNs)

PINNs represent the solution with a neural network and combine equation residuals, initial and boundary conditions, and observations in a common loss, supporting both forward simulation and parameter identification [41]. Variational and weak formulations use integral constraints to reduce reliance on pointwise high-order derivatives, while domain decomposition assigns local networks to different regions [28, 55, 26, 38]. Studies of gradient imbalance and neural tangent kernels explain why loss components can converge at different rates, motivating adaptive weighting and improved optimization [48, 50]. B-PINNs introduce Bayesian inference to quantify uncertainty in solutions and inferred parameters under noisy observations [53].

DeepONet and reduced-order operator learning

Operator learning trains a model across input functions so that new problem instances can be evaluated without solving each one from scratch. DeepONet combines a branch network encoding the input function with a trunk network encoding output coordinates to predict a continuous solution [36]. MIONet extends this construction to multiple input functions, while POD-DeepONet uses a reduced output basis extracted from solution data [27, 37]. PCA-Net compresses both input and output functions and learns a map between their coefficients [6]. Their accuracy depends on whether these finite representations preserve the main variations in the solution.

Spectral neural operators

Spectral architectures provide global interactions through transforms. The Fourier neural operator (FNO), a widely used architecture, learns Fourier multipliers and combines them with local mappings and nonlinear activations [32]. F-FNO factorizes spectral operations to reduce computational cost, whereas multiwavelet operators and WNO use spatial localization and scale decomposition to represent multiscale structure [45, 19, 46]. The Laplace neural operator (LNO) represents transient responses through poles and residues in the Laplace domain [8]. The choice of transform reflects the spatial and temporal structure of the solution.

Graph networks, geometry, and attention

Irregular meshes and varying domains require models that account for spatial geometry. MeshGraphNets learns local interactions and time evolution through mesh-based message passing; Geo-FNO and GINO connect physical domains to latent representations through learned deformations or graph operators [40, 31, 33]. Attention offers input-dependent interactions: Galerkin Transformer relates attention to integral approximation, while Transolver aggregates physical states to reduce the cost of interactions over large point sets [9, 51]. These methods support solution prediction beyond regular grids, including complex engineering geometries.

Physics-informed neural operators (PINO)

Physics-informed operator learning imposes governing equations across a family of inputs to reduce dependence on reference solutions. Early physics-constrained convolutional surrogates already explored parametric PDE learning without paired solution data [58]. PI-DeepONet incorporates equation and boundary residuals into DeepONet, while PINO combines data fitting with physics constraints, allowing coarse-grid solution data and fine-grid residual evaluation [49, 34]. Variational DeepONet and VINO instead formulate operator training through energy or weak-form objectives [17, 15]. This approach combines rapid operator evaluation with known physical laws, although residual discretization and loss balancing remain consequential training choices.

Kernel methods and random features

Kernel methods provide alternative representations for both PDE solutions and solution operators. Gaussian process methods impose PDE constraints in a kernel space, while operator kernel regression learns solution maps from finite observations of input and output functions [12, 4]. Random features reduce the representation to a finite coefficient problem: operator random-feature models learn maps between function spaces, and the Random Feature Method combines local features with domain decomposition for PDE solving [39, 10]. Fitting expansion coefficients simplifies optimization, although matrix size, conditioning, and nonlinear iterations can still be costly.

Learning within numerical solvers

Hybrid methods retain numerical algorithms and learn components that are expensive or difficult to design. Data-driven discretizations estimate differential operators on coarse grids; Solver-in-the-Loop includes numerical evolution during training; and HINTS combines neural operators with classical relaxation to reduce complementary error components [1, 47, 56]. For high-dimensional PDEs, deep backward schemes and deep splitting use stochastic representations and time discretization to obtain sequences of neural regression problems [25, 5]. The network thus learns a specific numerical subproblem within an established solution procedure.

Generative models for PDEs

Generative methods model distributions of solution fields, supporting inference from partial observations and uncertain data. DiffusionPDE combines diffusion priors with observation and physics constraints, while physics-informed diffusion models incorporate governing equations directly into training [24, 3]. PDE-Refiner uses iterative denoising to improve long autoregressive rollouts [35]. These approaches extend prediction to conditional sampling and iterative refinement, with additional computational costs from repeated generation steps.

PDE pretraining and foundation models

Recent work pretrains shared models on multiple equations and physical datasets before adapting them to specific tasks. DPOT and Poseidon study representation learning and transfer across PDEs, while the PDEformer family also encodes equations through symbolic or computational graph representations [21, 23, 54]. ICON uses input-output function pairs as context, enabling adaptation to new operator tasks through examples [52]. These models seek representations that can be reused across equations, conditions, and data distributions.

The Segal-Bargmann Transform

In this section, we introduce the Segal-Bargmann transform that will be used later. For more details we refer to [16 Chapter 1]. Throughout this article, our domain is either Rn\mathbb{R}^{n} or open subsets of Rn\mathbb{R}^{n}. We also assume the familiarity of the standard L2L^{2}-type Sobolev spaces Hs(Rn;R)H^{s}(\mathbb{R}^{n};\mathbb{R}). We will repeatedly use the standard Gamma function Γ(s)\Gamma(s), the incomplete upper and lower Gamma functions γ(s,x)\gamma(s,x) and Γ(s,x)\Gamma(s,x), respectively.

Precisely speaking, the Segal-Bargmann transform is a Gaussian-weighted Fourier transform with complex frequency. Throughout this article, we will interchangeably use Segal-Bargmann transform and aa-Segal-Bargmann transform. The latter one is used when we want to emphasize the importance of the parameter aa. All definitions below work equally well for complex-valued L2L^{2}-functions f∈L2(Rn;C)f\in L^{2}(\mathbb{R}^{n};\mathbb{C}).

Definition 3.1. Let L2(Rn;R)L^{2}(\mathbb{R}^{n};\mathbb{R}) be the collection of complex-valued square integrable measurable functions. We denote by the aa-Segal-Bargmann transform:

L2(Rn;R)∋f↦Ba(f)(z):=(2aπ)n4∫Rnexp⁡(−a2z⋅z−ax⋅x+2az⋅x)f(x) dx.(1)L^{2}(\mathbb{R}^{n};\mathbb{R})\ni f\mapsto B_{a}(f)(z):=\left(\frac{2a}{\pi}\right)^{\frac{n}{4}}\int_{\mathbb{R}^{n}}\exp\left(-\frac{a}{2}z\cdot z-ax\cdot x+2az\cdot x\right)f(x)\,dx. \tag*{(1)}

The function

Aa(z,x)=(2aπ)n4exp⁡(−a2z⋅z−ax⋅x+2az⋅x)A_{a}(z,x)=\left(\frac{2a}{\pi}\right)^{\frac{n}{4}}\exp\left(-\frac{a}{2}z\cdot z-ax\cdot x+2az\cdot x\right)

is called aa-Segal-Bargmann kernel.

It is constructed by Segal and Bargmann that the function Ba(f)B_{a}(f) is an entire function on Cn\mathbb{C}^{n}. Furthermore, it is sitting in a Hilbert space, called Bargmann-Fock space.

References

  1. [1]Yohai Bar-Sinai, Stephan Hoyer, Jason Hickey, and Michael P. Brenner. Learning data-driven discretizations for partial differential equations. Proceedings of the National Academy of Sciences, 116(31):15344–15349, 2019.
  2. [2]V. Bargmann. On a Hilbert space of analytic functions and an associated integral transform part I. Communications on Pure and Applied Mathematics, 14(3):187–214, 1961.DOI
  3. [3]Jan-Hendrik Bastek, WaiChing Sun, and Dennis M. Kochmann. Physics-Informed Diffusion Models. arXiv:2403.14404, 2024.arxiv.org/abs/2403.14404
  4. [4]Pau Balazs, Matthieu Darcy, Bamdad Hosseini, and Houman Owhadi. Kernel methods are competitive for operator learning. Journal of Computational Physics, 496:112549, 2024.
  5. [5]Christian Beck, Sebastian Becker, Patrick Cheridito, Arnulf Jentzen, and Ariel Neufeld. Deep Splitting Method for Parabolic PDEs. SIAM Journal on Scientific Computing, 43(5):A3135–A3154, 2021.DOI
  6. [6]Kaushik Bhattacharya, Bamdad Hosseini, Nikola B. Kovachki, and Andrew M. Stuart. Model Reduction And Neural Networks For Parametric PDEs. The SMAI Journal of computational mathematics, 7:121–157, 2021.arxiv.org/abs/2005.03180
  7. [7]E. Candès, J. Romberg, and T. Tao. Robust uncertainty principles: exact signal reconstruction from highly incomplete frequency information. IEEE Transactions on Information Theory, 52:489–509, 2006.arxiv.org/abs/math/0409186
  8. [8]Qianying Cao, Somdatta Goswami, and George Em Karniadakis. Laplace neural operator for solving differential equations. Nature Machine Intelligence, 6(6):631–640, 2024.DOI
  9. [9]Shuhao Cao. Choose a Transformer: Fourier or Galerkin. arXiv:2105.14995, 2021.arxiv.org/abs/2105.14995
  10. [10]Jingrun Chen, Xurong Chi, Weinan E, and Zhouwang Yang. Bridging Traditional and Machine Learning-based Algorithms for Solving PDEs: The Random Feature Method. arXiv:2207.13380, 2022.arxiv.org/abs/2207.13380
  11. [11]Tianping Chen and Hong Chen. Universal approximation to nonlinear operators by neural networks with arbitrary activation functions and its application to dynamical systems. IEEE Transactions on Neural Networks, 6(4):911–917, 1995.
  12. [12]Yifan Chen, Bamdad Hosseini, Houman Owhadi, and Andrew M. Stuart. Solving and learning nonlinear PDEs with Gaussian processes. Journal of Computational Physics, 447:110668, 2021.arxiv.org/abs/2103.12959
  13. [13]A. Cohen and G. Migliorati. Optimal weighted least-squares methods. SMAI J. Comput. Math., 3:181–231, 2017.arxiv.org/abs/1608.00512
  14. [14]Simon S. Du, Xiyu Zhai, Barnabás Póczos, and Aarti Singh. Gradient descent provably optimizes over-parameterized neural networks. In International Conference on Learning Representations, 2019.arxiv.org/abs/1810.02054
  15. [15]Mohammad Sadegh Eshaghi, Cosmin Anitescu, Manish Thombre, Yizheng Wang, Xiaoying Zhuang, and Timon Rabczuk. Variational Physics-informed Neural Operator (VINO) for solving partial differential equations. Computer Methods in Applied Mechanics and Engineering, 437:117785, 2025.
  16. [16]Gerald B. Folland. Harmonic Analysis in Phase Space, volume 122 of Annals of Mathematics Studies. Princeton University Press, Princeton, NJ, 1989.
  17. [17]Somdatta Goswami, Minglang Yin, Yue Yu, and George Em Karniadakis. A physics-informed variational DeepONet for predicting crack path in quasi-brittle materials. Computer Methods in Applied Mechanics and Engineering, 391:114587, 2022.DOI
  18. [18]Karlheinz Gröchenig. Foundations of Time-Frequency Analysis. Applied and Numerical Harmonic Analysis. Birkhäuser Boston, 2001.DOI
  19. [19]Gaurav Gupta, XionGye Xiao, and Paul Bogdan. Multiwavelet-based Operator Learning for Differential Equations. arXiv:2109.13459, 2021.arxiv.org/abs/2109.13459
  20. [20]Brian C. Hall. Holomorphic methods in analysis and mathematical physics. In First Summer School in Analysis and Mathematical Physics, volume 260 of Contemporary Mathematics, pages 1–59. American Mathematical Society, 2000.DOI
  21. [21]Zhongkai Hao, Chang Su, Songming Liu, Julius Berner, Chengyang Ying, Hang Su, Anima Anandkumar, Jian Song, and Jun Zhu. DPOT: Auto-Regressive Denoising Operator Transformer for Large-Scale PDE Pre-Training. arXiv:2403.03542, 2024.arxiv.org/abs/2403.03542
  22. [22]Zhongkai Hao, Zhengyi Wang, Hang Su, Chengyang Ying, Yinpeng Dong, Songming Liu, Ze Cheng, Jian Song, and Jun Zhu. GNOT: A General Neural Operator Transformer for Operator Learning. arXiv:2302.14376, 2023.arxiv.org/abs/2302.14376
  23. [23]Maximilian Herde, Bogdan Raonic, Tobias Rohner, Roger Käppeli, Roberto Molinaro, Emmanuel de Bézenac, and Siddhartha Mishra. Poseidon: Efficient Foundation Models for PDEs. arXiv:2405.19101, 2024.arxiv.org/abs/2405.19101
  24. [24]Jiahe Huang, Guandao Yang, Zichen Wang, and Jeong Joon Park. DiffusionPDE: Generative PDE-Solving Under Partial Observation. arXiv:2406.17763, 2024.arxiv.org/abs/2406.17763
  25. [25]Côme Huré, Huyền Pham, and Xavier Warin. Deep backward schemes for high-dimensional nonlinear PDEs. Mathematics of Computation, 89(324):1547–1579, 2020.
  26. [26]Ameya D. Jagtap and George Em Karniadakis. Extended Physics-Informed Neural Networks (XPINNs): A Generalized Space-Time Domain Decomposition Based Deep Learning Framework for Nonlinear Partial Differential Equations. Communications in Computational Physics, 28(5):2002–2041, 2020.
  27. [27]Pengzhan Jin, Shuai Meng, and Lu Lu. MIONet: Learning Multiple-Input Operators via Tensor Product. SIAM Journal on Scientific Computing, 44(6):A3490–A3514, 2022.arxiv.org/abs/2202.06137
  28. [28]E. Kharazmi, Z. Zhang, and G. E. Karniadakis. Variational Physics-Informed Neural Networks For Solving Partial Differential Equations. arXiv:1912.00873, 2019.arxiv.org/abs/1912.00873
  29. [29]Nikola Kovachki, Zongyi Li, Burigede Liu, Kamyar Aziz zaddehnesheli, Kaushik Bhattacharya, Andrew Stuart, and Anima Anandkumar. Neural Operator: Learning Maps Between Function Spaces. arXiv:2108.08481, 2021.DOI
  30. [30]B. Krötz, S. Thangavelu, and Y. Xu. The heat kernel transform for the Heisenberg group. J. Funct. Anal., 225:301–336, 2005.DOI
  31. [31]Zongyi Li, Daniel Zhengyu Huang, Burigede Liu, and Anima Anandkumar. Fourier Neural Operator with Learned Deformations for PDEs on General Geometries. arXiv:2207.05209, 2022.DOI
  32. [32]Zongyi Li, Nikola Kovachki, Kamyar Aziz zaddehnesheli, Burigede Liu, Kaushik Bhattacharya, Andrew Stuart, and Anima Anandkumar. Fourier Neural Operator for Parametric Partial Differential Equations. arXiv:2010.08895, 2020.arxiv.org/abs/2010.08895
  33. [33]Zongyi Li, Nikola Borislavov Kovachki, Chris Choy, Boyi Li, Jean Kossaifi, Shourya Prakash Otta, Mohammad Amin Nabian, Maximilian Stadler, Christian Hundt, Kamyar Aziz zaddehnesheli, and Anima Anandkumar. Geometry-Informed Neural Operator for Large-Scale 3D PDEs. arXiv:2309.00583, 2023.arxiv.org/abs/2309.00583
  34. [34]Zongyi Li, Hongkai Zheng, Nikola Kovachki, David Jin, Haoxuan Chen, Burigede Liu, Kamyar Aziz zaddehnesheli, and Anima Anandkumar. Physics-Informed Neural Operator for Learning Partial Differential Equations. ACM / IMS Journal of Data Science, 1(3):1–27, 2024.arxiv.org/abs/2111.03794
  35. [35]Philip Lippe, Bastiaan S. Veeling, Paris Perdikaris, Richard E. Turner, and Johannes Brandstetter. PDE-Refiner: Achieving Accurate Long Rollouts with Neural PDE Solvers. arXiv:2308.05732, 2023.arxiv.org/abs/2308.05732
  36. [36]Lu Lu, Pengzhan Jin, Guofei Pang, Zhongqiang Zhang, and George Em Karniadakis. Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators. Nature Machine Intelligence, 3(3):218–229, 2021.
  37. [37]Lu Lu, Xuhui Meng, Shengze Cai, Zhiping Mao, Somdatta Goswami, Zhongqiang Zhang, and George Em Karniadakis. A comprehensive and fair comparison of two neural operators (with practical extensions) based on FAIR data. Computer Methods in Applied Mechanics and Engineering, 393:114778, 2022.DOI
  38. [38]Ben Moseley, Andrew Markham, and Tarje Nissen-Meyer. Finite basis physics-informed neural networks (FBPINNs): a scalable domain decomposition approach for solving differential equations. Advances in Computational Mathematics, 49(4), 2023.DOI
  39. [39]Nicholas H. Nelsen and Andrew M. Stuart. The Random Feature Model for Input-Output Maps between Banach Spaces. SIAM Journal on Scientific Computing, 43(5):A3212–A3243, 2021.DOI
  40. [40]Tobias Pfaff, Meire Fortunato, Alvaro Sanchez-Gonzalez, and Peter W. Battaglia. Learning Mesh-Based Simulation with Graph Networks. arXiv:2010.03409, 2020.arxiv.org/abs/2010.03409
  41. [41]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, 378:686–707, 2019.DOI
  42. [42]Irving E. Segal. Mathematical characterization of the physical vacuum for a linear Bose–Einstein field. Illinois Journal of Mathematics, 6(3):500–523, 1962.DOI
  43. [43]K. Seip and R. Wallstén. Density theorems for sampling and interpolation in the Bargmann-Fock space II. Journal für die reine und angewandte Mathematik, 429:107–114, 1992.DOI
  44. [44]S. Thangavelu. Lectures on Hermite and Laguerre Expansions, volume 42 of Mathematical Notes. Princeton University Press, Princeton, New Jersey, 1993.
  45. [45]Alasdair Tran, Alexander Mathews, Lexing Xie, and Cheng Soon Ong. Factorized Fourier Neural Operators. arXiv:2111.13802, 2021.arxiv.org/abs/2111.13802
  46. [46]Tapas Tripura and Souvik Chakraborty. Wavelet Neural Operator for solving parametric partial differential equations in computational mechanics problems. Computer Methods in Applied Mechanics and Engineering, 404:115783, 2023.DOI
  47. [47]Kiwon Um, Robert Brand, Yun, Fei, Philipp Holl, and Nils Thuerey. Solver-in-the-Loop: Learning from Differentiable Physics to Interact with Iterative PDE-Solvers. arXiv:2007.00016, 2020.arxiv.org/abs/2007.00016
  48. [48]Sifan Wang, Yujun Teng, and Paris Perdikaris. Understanding and Mitigating Gradient Flow Pathologies in Physics-Informed Neural Networks. SIAM Journal on Scientific Computing, 43(5):A3055–A3081, 2021.
  49. [49]Sifan Wang, Hanwen Wang, and Paris Perdikaris. Learning the solution operator of parametric partial differential equations with physics-informed DeepONets. Science Advances, 7(40), 2021.
  50. [50]Sifan Wang, Xinling Yu, and Paris Perdikaris. When and why PINNs fail to train: A neural tangent kernel perspective. Journal of Computational Physics, 449:110768, 2022.arxiv.org/abs/2007.14527
  51. [51]Haixu Wu, Huakun Luo, Haowen Wang, Jianmin Wang, and Mingsheng Long. Transolver: A Fast Transformer Solver for PDEs on General Geometries. arXiv:2402.02366, 2024.arxiv.org/abs/2402.02366
  52. [52]Liu Yang, Siting Liu, Tingwei Meng, and Stanley J. Osher. In-context operator learning with data prompts for differential equation problems. Proceedings of the National Academy of Sciences, 120(39), 2023.DOI
  53. [53]Liu Yang, Xuhui Meng, and George Em Karniadakis. B-PINNs: Bayesian physics-informed neural networks for forward and inverse PDE problems with noisy data. Journal of Computational Physics, 425:109913, 2021.DOI
  54. [54]Zhanhong Ye, Zining Liu, Bingyang Wu, Hongjie Jiang, Leheng Chen, Minyan Zhang, Xiang Huang, Qinghe Meng. Jingyuan Zou, Hongsheng Liu, and Bin Dong. PDEformer-2: A Versatile Foundation Model for Two-Dimensional Partial Differential Equations. arXiv:2507.15409, 2025.arxiv.org/abs/2507.15409
  55. [55]Yaohua Zhang, Gang Bao, Xiaojing Ye, and Haomin Zhou. Weak adversarial networks for high-dimensional partial differential equations. Journal of Computational Physics, 411:109409, 2020.
  56. [56]Enrui Zhang, Adar Kahana, Alena Kopaničáková, Eli Turkel, Rishikesh Ranade, Jay Pathak, and George Em Karniadakis. Blending neural operators and relaxation methods in PDE numerical solvers. Nature Machine Intelligence, 6(11):1303–1313, 2024.arxiv.org/abs/2208.13273
  57. [57]Zecheng Zhang, Leung Wing Tat, and Hayden Schaeffer. BelNet: basis enhanced learning, a mesh-free neural operator. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 479(2276), 2023.arxiv.org/abs/2212.07336
  58. [58]Yinhao Zhu, Nicholas Zabaras, Phaedon-Stelios Koutsourelakis, and Paris Perdikaris. Physics-constrained deep learning for high-dimensional surrogate modeling and uncertainty quantification without labeled data. Journal of Computational Physics, 394:56–81, 2019.DOI

Paper details

vlnc:W_ecq22tvj
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC0-1.0
Usage

views · 10

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents