Struct PhysicsNeuralNet 

pub struct PhysicsNeuralNet {
    pub solver: GpuNeuralSolver,
    pub pde_weight: f64,
    pub bc_weight: f64,
}

Physics-informed neural network (PINN).

Wraps a GpuNeuralSolver and provides a composite loss that combines a PDE residual term with a boundary-condition MSE term.

Fields

solver: GpuNeuralSolver

Underlying neural solver.

pde_weight: f64

Weight applied to the PDE residual in the total loss.

bc_weight: f64

Weight applied to the boundary-condition loss in the total loss.

Implementations

impl PhysicsNeuralNet

pub fn new(layer_sizes: &[usize], pde_weight: f64, bc_weight: f64) -> Self

Create a new PINN from layer sizes, PDE weight, and BC weight.

pub fn total_loss(&self, pde_residual: f64, bc_loss: f64) -> f64

Compute the weighted total loss: pde_weight * |pde_residual| + bc_weight * bc_loss.

pub fn predict(&self, x: f64) -> f64

Predict the solution at a 1-D coordinate x.

Wraps the scalar input into a slice and extracts the first output.

Trait Implementations

impl Clone for PhysicsNeuralNet

fn clone(&self) -> PhysicsNeuralNet

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for PhysicsNeuralNet

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.