optim.PRIDU

optim.PRIDU#

class bartorch.optim.PRIDU#

Bases:

Primal-dual iteration, looping PRIDUBlock.

Parameters:

  • regularizers (Regularizer or ImplicitPrior, or an iterable of them, default=None) – Terms with auxiliary variables extend the optimization variable; the step spans the image and the auxiliary fields behind it, as in ADMM.

  • maxiter (int, default=30)

  • step (float, default=0.95) – Step size.

  • sigma_tau_ratio (float, default=1.0) – Ratio of the dual to the primal step: sigma = sqrt(step) * ratio, tau = sqrt(step) / ratio. BART’s own reconstructions set it to the factor the data was divided by, so pass data_scaling()’s value to match them.

  • adaptive_step (bool, default=False) – Adapt the steps during the iteration.

  • eigen (bool, default=False) – Scale the step by the largest eigenvalue of the normal operator, estimated with 30 power iterations.

  • hogwild (bool, default=False) – Decay the steps by a factor of 0.95 per iteration.

  • cclambda (float, default=0.0) – Weight of an identity added to the normal operator.

  • precond (LinearOperator, default=None) – Left preconditioner, lsqr2_create’s precond_op: chained onto the normal operator and onto the adjoint, so the iteration sees M(A^H A + lambda) x = M A^H y. Must be positive definite – BART composes it without symmetrizing. BART’s own reconstructions pass none.

__call__()#

Solve for the image given data y and encoding A.

Parameters:

  • y (torch.Tensor) – Data of A.oshape, recorded for autograd when it requires a gradient.

  • A (LinearOperator) – The encoding. A BART-backed operator is applied without leaving the library; a Python-defined one is called back once per application.

  • x0 (torch.Tensor, default=None) – Warm start of A.ishape; without one the iteration starts at zero.

Returns:

Complex64 solution of A.ishape, with y’s batch in front. A batch is solved one item at a time, each as its own run.

Return type:

torch.Tensor