cr.sparse.pursuit.mp.matrix_solve

cr.sparse.pursuit.mp.matrix_solve(Phi, y, max_iters=100, res_norm_rtol=0.0001, tracker=<function noop_tracker>)

Solves the sparse recovery problem \(y = \Phi x + e\) using matching pursuit algorithm

Parameters

• Phi (jax.numpy.ndarray) – A sensing matrix / dictionary

• y (jax.numpy.ndarray) – Measurements

• max_iters (int) – Maximum number of iterations

• res_norm_rtol (float) – Acceptable residual norm

Returns

A named tuple containing matching pursuit solution state

Return type

MPState

MPState interface is similar to cr.sparse.pursuit.RecoverySolution.

Examples

• Matching Pursuit

Note

If the expected sparsity of the solution is known then, one should set max_iters as about 4 times that value.