sift.kernels.MappingKernel#

class sift.kernels.MappingKernel(ignore_self=False, **kwargs)[source]#

Mapping kernel.

Define a kernel based on a given mapping \(T\).

\[k(x_i, x_j) = \sum_k p_{l}(k) p_{c}(k)\]

where \(p_{l} = T / T.sum(0), p_{c} = T / T.sum(1)\)

Parameters:

Attributes table#

`backend`	The kernel's `pykeops` backend.
`device`	The kernel's device.
`dtype`	The kernel's data type.
`k`	The instantiated kernel object.

MappingKernel.backend#

The kernel’s pykeops backend.

Can be either torch or numpy.