Skip to contents

The covariance calculation uses a matrix-factor Procrustes identity to avoid subtracting nearly equal covariance traces. The answer concerns the Gaussian distributions themselves, not an empirical sample from them.

Usage

ot_gaussian_distance(x, y)

Arguments

x, y

Objects returned by [ot_gaussian()].

Value

A `t4transport_comparison` with Wasserstein distance and squared cost. No finite transport plan is constructed.