Struct MultiAssetGbmProcess 

pub struct MultiAssetGbmProcess {
    pub drift_rates: Vec<f64>,
    pub vols: Vec<f64>,
    pub chol: Vec<Vec<f64>>,
}

N correlated lognormal assets as one N-state, N-factor process:

dS_i = (r - q_i) S_i dt + sigma_i S_i dW_i,   d<W_i, W_j> = rho_ij dt

The correlation enters through the rows of the lower-triangular Cholesky factor, so dw carries independent increments (the StochasticProcess contract). evolve overrides the generic Euler with the exact per-asset lognormal transition — under constant coefficients the joint law is exact at any step size, so coarse grids only cost monitoring resolution, never bias.

Fields

drift_rates: Vec<f64>

Per-asset risk-neutral drift rates r - q_i.

vols: Vec<f64>

chol: Vec<Vec<f64>>

Lower-triangular Cholesky factor of the asset correlation matrix.

Trait Implementations

impl StochasticProcess for MultiAssetGbmProcess

fn dim(&self) -> usize

Number of state variables.

fn factors(&self) -> usize

Number of driving Brownian factors (dw.len()).

fn drift(&self, _t: f64, x: &[f64], out: &mut [f64])

Drift vector a(t, x) into out (dim long).

fn diffusion(&self, _t: f64, x: &[f64], out: &mut [f64])

Diffusion matrix b(t, x) into out, row-major dim × factors.

fn evolve(&self, _t: f64, x: &[f64], dt: f64, dw: &[f64], out: &mut [f64])

One Euler-Maruyama step from x into out. The default allocates small scratch buffers; hot loops should override with model-specific stepping (which is also where non-Euler schemes — exact transitions, Heston full-truncation/QE — live, since generic multi-factor Milstein would need Lévy areas).

fn constrain(&self, _x: &mut [f64])

State constraint applied after every step (identity by default).