bartorch.casorati

Contents

bartorch.casorati#

bartorch.casorati()#

Casorati matrix of the overlapping blocks of input.

A block spans kernel_shape along axes and the whole of every other axis; the blocks are all its positions, with no wrapping.

Parameters:

  • input (torch.Tensor) – Array the blocks are taken from.

  • kernel_shape (int or tuple of int) – Block size, one per entry of axes.

  • axes (int or tuple of int) – Axes the blocks slide along.

Returns:

Shape (block_size, n_blocks): BART’s [blocks, block elements] matrix read in C order. Column s is block s flattened in C order, and blocks are numbered in C order of their start index.

Return type:

torch.Tensor

Contents