Physics Informed

90 public types in this namespace, organized by kind.

Models & Types (47)

Type Summary
AdvectionDiffusionEquation<T> Represents the Advection-Diffusion Equation: ∂c/∂t + v·∇c = D∇²c + S
AllenCahnEquation<T> Represents the Allen-Cahn equation: u_t - epsilon^2 * u_xx + u^3 - u = 0.
BlackScholesEquation<T> Represents the Black-Scholes Equation for option pricing: ∂V/∂t + ½σ²S²∂²V/∂S² + rS∂V/∂S - rV = 0
BurgersEquation<T> Represents the Burgers’ Equation: ∂u/∂t + u * ∂u/∂x = ν * ∂²u/∂x²
DeepOperatorNetwork<T> Implements Deep Operator Network (DeepONet) for learning operators.
DeepRitzMethod<T> Implements the Deep Ritz Method for solving variational problems and PDEs.
DirichletBoundaryCondition<T> Pins the solution to a fixed value on the boundary: u = value.
DomainDecompositionPINN<T> Domain Decomposition Physics-Informed Neural Network for large-scale problems.
DomainDecompositionTrainingHistory<T> Training history for domain decomposition PINN training.
FourierNeuralOperator<T> Implements the Fourier Neural Operator (FNO) for learning operators between function spaces.
GpuPINNTrainer<T> Provides GPU-accelerated training for Physics-Informed Neural Networks.
GpuTrainingHistory<T> Training history with GPU-specific metrics.
GraphNeuralOperator<T> Implements Graph Neural Operators for learning operators on graph-structured data.
HamiltonianNeuralNetwork<T> Implements Hamiltonian Neural Networks (HNN) for learning conservative dynamical systems.
HeatConductivityInverseProblem<T> Recovers an unknown thermal conductivity from measurements: given where and when you probed, and what you read, find the coefficient that explains it.
HeatEquation<T> Represents the Heat Equation (or Diffusion Equation): ∂u/∂t = α ∂²u/∂x²
InterfaceDefinition<T> Defines an interface between two subdomains.
InverseProblemPINN<T> Implements a Physics-Informed Neural Network for inverse problems (parameter identification).
InverseProblemResult<T> Results from inverse problem optimization.
KortewegDeVriesEquation<T> Represents the Korteweg-de Vries (KdV) Equation: ∂u/∂t + αu∂u/∂x + β∂³u/∂x³ = 0
LagrangianNeuralNetwork<T> Implements Lagrangian Neural Networks (LNN) for learning mechanical systems.
LinearElasticityEquation<T> Represents the 2D Linear Elasticity Equations (Navier-Cauchy equations): (λ + μ)∂(∂u/∂x + ∂v/∂y)/∂x + μ∇²u + fₓ = 0 (λ + μ)∂(∂u/∂x + ∂v/∂y)/∂y + μ∇²v + fᵧ = 0
MaxwellEquations<T> Represents Maxwell’s equations for electromagnetic wave propagation (2D TE mode).
MultiFidelityPINN<T> Multi-fidelity Physics-Informed Neural Network for combining data of different accuracy levels.
MultiFidelityTrainingHistory<T> Training history for multi-fidelity PINN training.
MultiScalePINN<T> Implements a Multi-Scale Physics-Informed Neural Network for solving PDEs with multiple scales.
MultiScaleReactionDiffusion<T> A two-scale reaction-diffusion equation: the same physics at a coarse and a fine length scale, with a coupling term that penalises the two disagreeing where they overlap.
NavierStokesEquation<T> Represents the incompressible Navier-Stokes equations for fluid dynamics.
NeumannBoundaryCondition<T> Pins the solution’s slope on the boundary rather than its value: ∂u/∂n = flux.
OperatorBenchmarkResult
OperatorDataset2D
PDEDerivatives<T> Holds the derivatives needed for PDE computation.
PDEResidualGradient<T> Holds gradients of a residual with respect to outputs and derivatives.
PINNGpuMemoryInfo GPU memory usage information for PINN training.
PdeBenchmarkResult
PhysicsInformedLoss<T> Loss function for Physics-Informed Neural Networks (PINNs).
PhysicsInformedNeuralNetwork<T> Represents a Physics-Informed Neural Network (PINN) for solving PDEs.
PhysicsLossGradient<T> Holds loss and gradient information for physics-informed objectives.
PoissonEquation<T> Represents the Poisson Equation: ∇²u = f(x,y)
SchrodingerEquation<T> Represents the time-dependent Schrodinger equation for quantum mechanics.
SubdomainDefinition<T> Defines a subdomain for domain decomposition.
SymbolicExpression<T> Represents a symbolic mathematical expression.
SymbolicPhysicsLearner<T> Implements Symbolic Physics Learning for discovering interpretable equations from data.
TrainingHistory<T> Stores training history for analysis.
UniversalDifferentialEquation<T> Implements Universal Differential Equations (UDEs) - ODEs with neural network components.
VariationalPINN<T> Implements Variational Physics-Informed Neural Networks (VPINNs).
WaveEquation<T> Represents the Wave Equation: ∂²u/∂t² = c² * ∇²u

Layers (1)

Type Summary
FourierLayer<T> Represents a single Fourier layer in the FNO.

Base Classes (1)

Type Summary
PDESpecificationBase<T> Base class for all Partial Differential Equation (PDE) specifications.

Interfaces (12)

Type Summary
IBoundaryConditionGradient<T> Provides gradients for boundary residuals with respect to outputs and derivatives.
IBoundaryCondition<T> Defines boundary conditions for a PDE problem.
IDomainDecompositionTrainingHistory<T> Extended training history interface for domain decomposition PINN training.
IGpuAcceleratedPINN<T> Interface for Physics-Informed Neural Networks that support GPU acceleration.
IInitialCondition<T> Defines initial conditions for time-dependent PDEs.
IInverseProblemGradient<T> Provides gradient information for inverse problem parameters.
IInverseProblem<T> Defines the interface for inverse problems in physics-informed neural networks.
IMultiFidelityTrainingHistory<T> Extended training history interface for multi-fidelity PINN training.
IMultiScalePDEGradient<T> Provides gradient information for multi-scale PDE residuals.
IMultiScalePDE<T> Defines the interface for multi-scale Partial Differential Equations.
IPDEResidualGradient<T> Provides gradients of the PDE residual with respect to outputs and derivatives.
IPDESpecification<T> Defines the interface for specifying Partial Differential Equations (PDEs) that can be used with Physics-Informed Neural Networks.

Enums (3)

Type Summary
InverseProblemRegularization Specifies the type of regularization for inverse problems.
OdeIntegrationMethod Supported integration schemes for UDE simulation.
SymbolicExpressionType Types of symbolic expressions.

Options & Configuration (21)

Type Summary
AllenCahnBenchmarkOptions The Allen-Cahn benchmark, u_t = ε² u_xx + u - u³, a reaction-diffusion equation whose solution separates into phases at ±1 joined by thin interfaces.
BurgersBenchmarkOptions The viscous Burgers benchmark, u_t + u u_x = ν u_xx, the shock-forming equation of Raissi et al.
DarcyOperatorBenchmarkOptions Configures the 2D Darcy-flow operator-learning dataset that OperatorBenchmarkSuite generates: each sample pairs a random permeability field a(x, y) with the pressure u solving -∇·(a∇u) = f on the unit square.
DeepOperatorNetworkOptions Configuration options for the DeepOperatorNetwork.
DeepRitzMethodOptions Configuration options for the DeepRitzMethod.
DomainDecompositionPINNOptions Configuration options for the DomainDecompositionPINN.
FourierNeuralOperatorOptions Configuration options for the FourierNeuralOperator.
GpuPINNTrainingOptions Configuration options for GPU-accelerated PINN training.
GraphNeuralOperatorOptions Configuration options for the GraphNeuralOperator.
HamiltonianNeuralNetworkOptions Configuration options for the HamiltonianNeuralNetwork.
InverseProblemOptions<T> Configuration options for inverse problem PINN training.
LagrangianNeuralNetworkOptions Configuration options for the LagrangianNeuralNetwork.
MultiFidelityPINNOptions Configuration options for the MultiFidelityPINN.
MultiScalePINNOptions Configuration options for the Multi-Scale PINN model.
MultiScaleTrainingOptions<T> Configuration options for multi-scale PINN training.
OperatorBenchmarkOptions Configures the 1D smoothing-operator benchmark of OperatorBenchmarkSuite: each input is a random sine series on [0, 1], and the target is its moving average, which a predictor is scored against.
PdeBenchmarkOptions The grid and time span shared by the 1D time-dependent PDE benchmarks of PdeBenchmarkSuite: a finite-difference reference solution is computed on this grid and a predictor u(x, t) is scored against it at FinalTime.
PhysicsInformedNeuralNetworkOptions Configuration options for the PhysicsInformedNeuralNetwork.
PoissonOperatorBenchmarkOptions Configures the 2D Poisson operator-learning dataset that OperatorBenchmarkSuite generates: each sample pairs a random forcing field f(x, y) with the u solving -Δu = f on the unit square.
UniversalDifferentialEquationsOptions Configuration options for the UniversalDifferentialEquations model.
VariationalPINNOptions Configuration options for the VariationalPINN.

Helpers & Utilities (5)

Type Summary
DeepRitzEnergy<T> Standard energy densities for DeepRitzMethod, batched over the quadrature points.
NeuralNetworkDerivatives<T> Exact first and second derivatives of a network’s outputs with respect to its inputs, for physics-informed losses (gradients, Laplacians, Hessians of the PDE residual).
OperatorBenchmarkSuite
PdeBenchmarkSuite
VariationalWeakForm<T> Standard weak-form integrands for VariationalPINN, batched over the quadrature points.