optim.ist

Contents

optim.ist#

bartorch.optim.ist()#

Solve a regularized least-squares problem by iterative soft thresholding.

Minimizes \(\tfrac12 \| A x - y \|^2 + g(x)\) by alternating a gradient step on the data-fidelity term with the proximal operator of g. Takes exactly one regularizer, whose linear transform must be the identity.

Parameters:

  • y (tensor) – Data of A.oshape.

  • A (LinearOperator) – The encoding operator.

  • regularizers (Regularizer or ImplicitPrior, default=None) – The single term g.

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

  • **settings – Settings of IST, among them maxiter (30), step and eigen.

Returns:

Complex64 solution of A.ishape.

Return type:

torch.Tensor

Contents