spyrit.core.noise.NoNoise.reindex

NoNoise.reindex(x: tensor, axis: str = 'rows', inverse_permutation: bool = False) → tensor

Sorts a tensor along a specified axis using the indices tensor. The indices tensor is contained in the attribute self.meas_op.indices.

The indices tensor contains the new indices of the elements in the values tensor. values[0] will be placed at the index indices[0], values[1] at indices[1], and so on.

Using the inverse permutation allows to revert the permutation: in this case, it is the element at index indices[0] that will be placed at the index 0, the element at index indices[1] that will be placed at the index 1, and so on.

Note

See reindex() for more details.

Args:

values (torch.tensor): The tensor to sort. Can be 1D, 2D, or any multi-dimensional batch of 2D tensors.

axis (str, optional): The axis to sort along. Must be either ‘rows’ or ‘cols’. If values is 1D, axis is not used. Default is ‘rows’.

inverse_permutation (bool, optional): Whether to apply the permutation inverse. Default is False.

Raises:

ValueError: If axis is not ‘rows’ or ‘cols’.

Returns:

torch.tensor: The sorted tensor by the given indices along the specified axis.