cr.sparse.cvx.adm.yall1.solve

cr.sparse.cvx.adm.yall1.solve(A, b, x0=None, z0=None, W=None, weights=None, nonneg=False, rho=0.0, delta=0.0, gamma=1.0, tolerance=0.005, max_iters=9999, jit=True)[source]

Wrapper method to solve a variety of l1 minimization problems using ADMM

Parameters

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

  • b (jax.numpy.ndarray) – Signal being approximated

  • x0 (jax.numpy.ndarray) – Initial value of solution (primary variable) \(x\)

  • z0 (jax.numpy.ndarray) – Initial value of dual variable \(z\)

  • nonneg (bool) – Flag to indicate if values in the solution are all non-negative

  • W (jax.numpy.ndarray) – The sparsifying orthonormal basis such that \(W x\) is sparse

  • weights (jax.numpy.ndarray) – The weights for individual entries in \(x\)

  • rho (float) – weight for the quadratic penalty term

  • delta (float) – constraint on the residual norm

  • gamma (float) – ADMM update parameter for \(x\)

  • max_iters (int) – maximum number of ADMM iterations

Returns

Solution vector \(x\) and residual \(r\)

Return type

RecoveryFullSolution

This function is based on [YZ11]. It implements eq 2.25 of the paper.