Distance#

Distance types#

#include <cuvs/distance/distance.h>

enum cuvsDistanceType#

enum to tell how to compute distance

Values:

enumerator L2Expanded#: evaluate as dist_ij = sum(x_ik^2) + sum(y_ij)^2 - 2*sum(x_ik * y_jk)

enumerator L2SqrtExpanded#: same as above, but inside the epilogue, perform square root operation

enumerator CosineExpanded#: cosine distance

enumerator L1#: L1 distance

enumerator L2Unexpanded#: evaluate as dist_ij += (x_ik - y-jk)^2

enumerator L2SqrtUnexpanded#: same as above, but inside the epilogue, perform square root operation

enumerator InnerProduct#: basic inner product

enumerator Linf#: Chebyshev (Linf) distance

enumerator Canberra#: Canberra distance

enumerator LpUnexpanded#: Generalized Minkowski distance

enumerator CorrelationExpanded#: Correlation distance

enumerator JaccardExpanded#: Jaccard distance

enumerator HellingerExpanded#: Hellinger distance

enumerator Haversine#: Haversine distance

enumerator BrayCurtis#: Bray-Curtis distance

enumerator JensenShannon#: Jensen-Shannon distance

enumerator HammingUnexpanded#: Hamming distance

enumerator KLDivergence#: KLDivergence

enumerator RusselRaoExpanded#: RusselRao

enumerator DiceExpanded#: Dice-Sorensen distance

enumerator BitwiseHamming#: Bitstring Hamming distance

enumerator Precomputed#: Precomputed (special value)

Pairwise distance#

#include <cuvs/distance/pairwise_distance.h>

cuvsError_t cuvsPairwiseDistance( cuvsResources_t res, DLManagedTensor *x, DLManagedTensor *y, DLManagedTensor *dist, cuvsDistanceType metric, float metric_arg )#

Compute pairwise distances for two matrices.

Usage example:

#include <cuvs/core/c_api.h>
#include <cuvs/distance/pairwise_distance.h>

// Create cuvsResources_t
cuvsResources_t res;
cuvsError_t res_create_status = cuvsResourcesCreate(&res);

// Assume a populated `DLManagedTensor` type here
DLManagedTensor x;
DLManagedTensor y;
DLManagedTensor dist;

cuvsPairwiseDistance(res, &x, &y, &dist, L2SqrtUnexpanded, 2.0);

Parameters:

res – [in] cuvs resources object for managing expensive resources
x – [in] first set of points (size n*k)
y – [in] second set of points (size m*k)
dist – [out] output distance matrix (size n*m)
metric – [in] distance to evaluate
metric_arg – [in] metric argument (used for Minkowski distance)