Parametric Gaussian probability measures — ot

Keep Gaussian parameters separate from finite samples. Univariate variance zero is supported as a degenerate point mass. Covariances in dimensions greater than one must be strictly positive definite. No simulation, covariance ridge, or PSD projection is performed.

Usage

ot_gaussian(mean, covariance)

Arguments

mean: A nonempty finite real mean vector.
covariance: A covariance matrix matching the mean dimension; a scalar variance is also accepted for a univariate Gaussian.

Value

A `t4transport_gaussian` object containing mean, covariance (`var`), dimension, and its covariance-domain declaration.

Examples

x <- ot_gaussian(0, 1)
y <- ot_gaussian(3, 4)
ot_gaussian_distance(x, y)
#> Transport comparison | gaussian_formula | euclidean | order 2 
#> Status: SUCCESS ( gaussian_parameter_formula )
#> Transport cost: 10 
#> Wasserstein distance: 3.162278