linop.Convolve

Contents

linop.Convolve#

class bartorch.linop.Convolve#

Bases: LinearOperator

Convolution with a fixed kernel, BART’s linop_conv.

Parameters:

  • kernel (tensor) – What to convolve with, with one axis per axis of shape; an axis it has only one of is left alone.

  • shape (tuple of int) – The domain, C order.

  • axes (int or tuple of int) – Which axes to convolve along.

  • mode (str, default='same') – How the ends are treated, under numpy’s names: "wrap" (circular), "same" (the input’s size, truncated), "valid" (only where the kernel fits) or "full" (extended).

  • direction (str, default=None) – "symmetric", "causal" or "anticausal". BART only plans a "valid" or "full" convolution as a causal one, so leaving this out picks causal for those two and symmetric for the others rather than failing inside BART’s planner.

Also has the methods and properties of LinearOperator.

Contents