spyrit.core.meas.Linear.adjoint

Linear.adjoint(y: tensor) → tensor

Applies adjoint transform to incoming measurements \(x = H^{T}y\)

This brings back the measurements in the image domain, but is not equivalent to the inverse of the forward operator.

Args:

\(y\) (torch.tensor): batch of measurement vectors of shape \((*, M)\) where * denotes any number of dimensions (e.g. (b,c) where b is the batch size and c the number of channels) and M the number of measurements.

Output:

torch.tensor: The adjoint of the input measurements, which are in the image domain. It has shape \((*, h, w)\) where * denotes any number of dimensions and h, w the height and width of the images.

Shape:

\(y\): \((*, M)\)

Output: \((*, h, w)\)

Example:

>>> H = torch.randn([400, 1600])
>>> meas_op = Linear(H)
>>> y = torch.randn([10, 400]
>>> x = meas_op.adjoint(y)
>>> print(x.shape)
torch.Size([10, 40, 40])