egttools.utils.calculate_stationary_distribution_non_hermitian

calculate_stationary_distribution_non_hermitian(transition_matrix)[source]

Calculates the stationary distribution of a Markov chain using a Schur decomposition.

Produces the same result as calculate_stationary_distribution(); the Schur decomposition is more numerically stable than a full eigendecomposition for ill-conditioned dense matrices.

The matrix must be column-stochastic (each column sums to 1), i.e. the transpose of the row-stochastic matrices returned by calculate_transition_matrix. Pass T.T. Sparse input is kept sparse and solved as in calculate_stationary_distribution().

Parameters:

transition_matrix (Union[ndarray, spmatrix, sparray]) – A 2 dimensional, column-stochastic transition matrix.

Returns:

A 1-dimensional vector containing the stationary distribution

Return type:

ndarray

Warns:

RuntimeWarning – If some transition probability is close to 1, or if eigenvalue 1 is not simple (the chain has no unique stationary distribution).