cr.sparse.pursuit.omp.matrix_solve_jit

cr.sparse.pursuit.omp.matrix_solve_jit(Phi, y, max_iters, max_res_norm=1e-06)

Solves the recovery/approximation problem \(y = \Phi x + e\) using Orthogonal Matching Pursuit

Parameters
  • Phi – A matrix representing the underdetermined linear system

  • y (jax.numpy.ndarray) – The signal to be modeled by OMP

  • max_iters (int) – Sparsity of the solution vector (number of nonzero entries)

Note

In order to support JIT compilation of this function, the halting criteria for residual norm has been ignored for now. The current implementation simply unrolls OMP main loop depending on max_iters. This may be revised in future.