sift.kernels.KnnKernel#

class sift.kernels.KnnKernel(kernel=None, dtype=None, device=None, backend=None)[source]#

k-NN kernel.

Defines a kernel based on a given pre-computed k-nn similarity and applies row-normalization to the provided kernel.

Attributes table#

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

KnnKernel.backend#

The kernel’s pykeops backend.

Can be either torch or numpy.