Struct Hnn 

pub struct Hnn { /* private fields */ }

A Hamiltonian network H_θ(q, p) (two inputs, one scalar output).

Implementations

impl Hnn

pub fn new(hidden: usize, seed: u64) -> Self

A Hamiltonian network with the given hidden width.

pub fn hamiltonian(&self, q: f64, p: f64) -> f64

The learned Hamiltonian (energy) at a phase point.

pub fn field(&self, q: f64, p: f64) -> (f64, f64)

The learned dynamics (q̇, ṗ) = (∂H/∂p, −∂H/∂q) at a phase point.

pub fn step_rk4(&self, q: f64, p: f64, dt: f64) -> (f64, f64)

One RK4 rollout step of the learned dynamics.

pub fn train_step( &mut self, q: &[f64], p: &[f64], qdot: &[f64], pdot: &[f64], lr: f64, ) -> f64

One Adam step matching the learned field (∂H/∂p, −∂H/∂q) to the observed (q̇, ṗ). Returns the loss.

pub fn train( &mut self, q: &[f64], p: &[f64], qdot: &[f64], pdot: &[f64], epochs: usize, lr: f64, ) -> f64

Train for epochs steps; returns the final loss.