sparse.greedy_pursuit.matching_pursuit

sparse.greedy_pursuit.matching_pursuit(mat_a, b, n_iters, weak_threshold=1.0, tol=1e-09)

(Weak) Matching Pursuit (MP, WMP) algorithms find an approximate solution to (1). Compared to OMP, MP and WMP algorithms are weaker (in terms of having larger residual) but faster.

Parameters

mat_a(N, M) np.ndarray

The input weight matrix \(\boldsymbol{A}\).

b(N,) np.ndarray

The right side of the equation (1).

n_itersint

The number of iterations to perform. The number of non-zero coefficients in the solution \(\vec{x}\) is at most n_iters.

weak_thresholdfloat, optional

A threshold in range (0, 1] for WMP algorithm that defines an early stop. If set to 1., MP algorithm is used. Default is 1. (MP).

tolfloat, optional

The tolerance which determines when a solution is “close enough” to the optimal solution. Compared with L2-norm of residuals (errors) at each iteration.

Returns

Solution

Refer to Greedy Matching Pursuit algorithms.

Notes

If the sparseness condition (1) is satisfied, the true unknown solution is recovered.