GAIA introduces a geometry-adaptive integral autoencoder that unifies forward, boundary-value, and inverse PDE operator learning on arbitrary domains via geometry tokens and cross-attention.
super hub Mixed citations
Fourier Neural Operator for Parametric Partial Differential Equations
Mixed citation behavior. Most common role is background (63%).
261
Pith papers citing it
Background
63% of classified citations
abstract
The classical development of neural networks has primarily focused on learning mappings between finite-dimensional Euclidean spaces. Recently, this has been generalized to neural operators that learn mappings between function spaces. For partial differential equations (PDEs), neural operators directly learn the mapping from any functional parametric dependence to the solution. Thus, they learn an entire family of PDEs, in contrast to classical methods which solve one instance of the equation. In this work, we formulate a new neural operator by parameterizing the integral kernel directly in Fourier space, allowing for an expressive and efficient architecture. We perform experiments on Burgers' equation, Darcy flow, and Navier-Stokes equation. The Fourier neural operator is the first ML-based method to successfully model turbulent flows with zero-shot super-resolution. It is up to three orders of magnitude faster compared to traditional PDE solvers. Additionally, it achieves superior accuracy compared to previous learning-based solvers under fixed resolution.
hub tools
citation-role summary
background 26
method 12
baseline 2
other 1
citation-polarity summary
claims ledger
- abstract The classical development of neural networks has primarily focused on learning mappings between finite-dimensional Euclidean spaces. Recently, this has been generalized to neural operators that learn mappings between function spaces. For partial differential equations (PDEs), neural operators directly learn the mapping from any functional parametric dependence to the solution. Thus, they learn an entire family of PDEs, in contrast to classical methods which solve one instance of the equation. In this work, we formulate a new neural operator by parameterizing the integral kernel directly in Fou
authors
co-cited works
representative citing papers
A systematic approach maps any-dimensional invariant functions to a unique function on an infinite-dimensional limit space admitting a topology with compact sets where universality holds, with examples of non-universal architectures and fixes.
KANs with learnable univariate spline activations on edges achieve better accuracy than MLPs with fewer parameters, faster scaling, and direct visualization for scientific discovery.
A self-explainable operator learning method reformulates operators as decomposable integral equations to reveal spatial input contributions to predictions in blood flow and aerodynamics problems.
PNOT combines graph attention on boundary heat flux with a physics-aware neural operator and gradient-constrained loss to reconstruct divertor temperature fields for real-time fusion control.
McMg is a learned phase-space multi-channel multigrid preconditioner that maps residuals to corrections for heterogeneous Helmholtz equations and shows fewer iterations than classical and neural baselines in tests.
HO-FNO extends standard FNO with n-linear spectral mixing and shows improved accuracy on nonlinear PDE benchmarks, sometimes with a single layer beating deeper FNO models.
DEN is an unsupervised neural framework that uses dimension expansion to enable efficient inverse design of nanophotonic structures from low-dimensional objectives via differentiable simulations.
Hybrid exact-learned equivariant operator for incompressible Stokes flow fixes the known core kernel exactly and learns only the boundary correction as a second-kind operator, achieving high accuracy, data efficiency, and cross-shape generalization.
HNO using real Hartley transform outperforms FNO on elliptic PDEs while FNO wins on time-dependent PDEs, yielding a rule to match spectral basis to operator symmetry via Green's functions.
A data-driven variational discretization of Onsager's principle learns uncertain free-energy and dissipation functionals from observations while guaranteeing provable energy stability for arbitrarily long simulations.
ELADO provides a benchmark suite of elliptic PDE datasets designed to isolate and quantify failure modes in neural operator architectures.
Optimizing training data via a differentiable SCM yields climate emulators that outperform those trained on six standard ScenarioMIP pathways while using less data and isolating distinct forcing responses.
Hybrid GNN-FEM surrogate replaces only the phase-field update step in a staggered incremental scheme for phase-field fracture, using dimensionless mesh features and physics-informed loss to generalize across geometries, loads, materials, and discretizations.
FEMONet learns parameterized optical scattering operators by mapping physical parameters to Galerkin-consistent finite-element solution spaces for complex-valued wave problems.
The authors combine H(div)-L2 subspaces from Raviart-Thomas and dgP0 elements with a transformer and GP regression on fluxes to create real-time structure-preserving surrogates with closed-form posterior uncertainty for Dirichlet-to-Neumann maps.
Proves a weighted Nachbin theorem establishing universal approximation of differentiable maps from weighted infinite-dimensional manifolds to Banach spaces, including derivatives, with applications to non-anticipative path functionals and signature methods.
d_eff in PINNs is shown to be an operator invariant equal to kernel dimension for finite-kernel operators, enabling subspace projection for physics-preserving constraint adaptation.
FreqNO-DPS corrects neural operator spectral bias in 3D elastic wavefield prediction by frequency-dependent guidance in diffusion posterior sampling conditioned on sparse observations, achieving near-zero bias at 2-5% sensor coverage.
TransportBench is a new high-fidelity dataset and standardized benchmark that evaluates neural architectures on non-equilibrium flow transport and finds performance depends strongly on specific flow characteristics.
Proposes residual-based physics-informed coarsening in multigrid GNNs to allocate capacity to high-activity regions for more stable solid mechanics surrogates.
GON uses 2-jet features and an anchor-and-variance objective to fix gauge freedom in ordinal predictability scoring, enabling pretrained initialization to outperform scratch training on held-out dynamical systems.
Therm-FM adapts a pretrained PDE foundation model using thermal-equivalent multi-fidelity training to achieve up to 10.6x lower error in 3D-IC thermal simulation with under 20% of typical training data and strong cross-design transfer.
A framework pretrained on authentic binary occlusion masks uses guided sampling and intersection-based partitioning to train diffusion models on incomplete physical observations without zero-query regions.
citing papers explorer
-
Implicit Neural Optimal Transport via Fixed-Point Optimization
A single-network implicit neural optimal transport method that solves the c-transform via proximal fixed-point iteration for stable, non-adversarial training.
-
Operator Learning for PDE Backstepping Control of Parabolic Equations on Time-Varying Domains
DeepONet learns the operator from moving-boundary trajectories to backstepping kernels for parabolic PDE stabilization on time-varying domains, yielding exponential decay on finite intervals and ~1000x speedup over repeated kernel PDE solves.
-
PI-SONet: A Physics-Informed Symplectic Operator Network for Real-Time Optimal Control of Multi-Agent Systems
PI-SONet trains a single structure-preserving operator network to deliver sub-second approximations to Pontryagin Maximum Principle solutions for parameterized multi-agent optimal control problems.
-
Man, Machine, and Mathematics
A high-level outline is given for a unified theory that reduces learning to a small set of ideas from dynamical systems, geometry, and physics via definitions of solvable problems and parametrized methods.