Neighbors Refine

Source header: cuvs/neighbors/refine.h

cuvsRefine

Refine nearest neighbor search.

1 CUVS_EXPORT cuvsError_t cuvsRefine(cuvsResources_t res,
2 DLManagedTensor* dataset,
3 DLManagedTensor* queries,
4 DLManagedTensor* candidates,
5 cuvsDistanceType metric,
6 DLManagedTensor* indices,
7 DLManagedTensor* distances);

Refinement is an operation that follows an approximate NN search. The approximate search has already selected n_candidates neighbor candidates for each query. We narrow it down to k neighbors. For each query, we calculate the exact distance between the query and its n_candidates neighbor candidate, and select the k nearest ones.

Parameters

Name	Direction	Type	Description
`res`	in	`cuvsResources_t`	cuvsResources_t opaque C handle
`dataset`	in	`DLManagedTensor*`	device matrix that stores the dataset [n_rows, dims]
`queries`	in	`DLManagedTensor*`	device matrix of the queries [n_queris, dims]
`candidates`	in	`DLManagedTensor*`	indices of candidate vectors [n_queries, n_candidates], where n_candidates >= k
`metric`	in	`cuvsDistanceType`	distance metric to use. Euclidean (L2) is used by default
`indices`	out	`DLManagedTensor*`	device matrix that stores the refined indices [n_queries, k]
`distances`	out	`DLManagedTensor*`	device matrix that stores the refined distances [n_queries, k]

Returns

CUVS_EXPORT cuvsError_t