optim.POCS#

class bartorch.optim.POCS#

Bases:

Projection onto convex sets, looping POCSBlock.

maxiter sweeps of the projections, with no stopping test, as BART’s pocs.

Parameters:

projections (sequence) – Callables on a tensor, or bartorch.priors terms, as POCSBlock takes them.
maxiter (int, default=50) – Number of sweeps; pocsense’s default.

Examples

>>> POCS([consistency, sense, sparsity], maxiter=50)(kspace)

__call__()#

Sweep the projections.

Parameters:

y (tensor) – Fixes the shape, dtype and device of the zero start; its values are not read.
A (object, default=None) – Ignored.
x0 (tensor, default=None) – Starting iterate; zero of y’s shape when omitted.

Returns:

The iterate after the last sweep.

Return type:

torch.Tensor

optim.POCS