riecovest.distance

Functions for computing distances between matrices.

Most of the functions are distances between positive definite matrices, although some are for general matrices. All operations are defined in jax, so that they can be used in algorithms requiring automatic differentiation.

Functions

airm(A, B)	The affine invariant Riemannian metric distance between two positive definite matrices.
corr_matrix_distance(A, B)	Computes the correlation matrix distance
frob_gevd_weighted_fullrank(A, B)	The frobenious distance between A and B, weighted by the generalized eigenvectors.
frob_sq(A, B)	The squared Frobenius distance between two matrices A and B.
frob_sq_weighted(A, B, W)	The squared Frobenius distance between two matrices A and B, weighted by a matrix W.
kl_divergence_gaussian(A, B)	The Kullback Leibler divergence between two Gaussian distributions
tyler_log_likelihood(data, cov)	Log likelihood function for a complex elliptically symmetric distribution.
tyler_log_likelihood_no_normalization(data, cov)	Log likelihood function for a complex elliptically symmetric distribution.
wasserstein_distance(A, B)	Computes the Wasserstein distance between two zero-mean Gaussian distributions defined by two positive definite matrices.
wishart_likelihood(mat_variable, cov, N)	Wishart likelihood function.
wishart_log_likelihood(mat_variable, cov, N)	Wishart log likelihood function.

Classes

partial(func, /, *args, **keywords)

Create a new function with partial application of the given arguments and keywords.