Attention

The vector search and clustering algorithms in RAFT are being migrated to a new library dedicated to vector search called cuVS. We will continue to support the vector search algorithms in RAFT during this move, but will no longer update them after the RAPIDS 24.06 (June) release. We plan to complete the migration by RAPIDS 24.08 (August) release.

Clustering Model Scoring#

Adjusted Rand Index#

#include <raft/stats/adjusted_rand_index.cuh>

namespace raft::stats

template<typename value_t, typename math_t, typename idx_t>
double adjusted_rand_index(raft::resources const &handle, raft::device_vector_view<const value_t, idx_t> first_cluster_array, raft::device_vector_view<const value_t, idx_t> second_cluster_array)#

Function to calculate Adjusted RandIndex.

See also

https://en.wikipedia.org/wiki/Rand_index

Template Parameters:

  • value_t – data-type for input label arrays

  • math_t – integral data-type used for computing n-choose-r

  • idx_t – Index type of matrix extent.

Parameters:

  • handle – [in] the raft handle.

  • first_cluster_array – [in] the array of classes

  • second_cluster_array – [in] the array of classes

Returns:

the Adjusted RandIndex

Completeness Score#

#include <raft/stats/completeness_score.cuh>

namespace raft::stats

template<typename value_t, typename idx_t>
double completeness_score(raft::resources const &handle, raft::device_vector_view<const value_t, idx_t> truth_cluster_array, raft::device_vector_view<const value_t, idx_t> pred_cluster_array, value_t lower_label_range, value_t upper_label_range)#

Function to calculate the completeness score between two clusters.

Template Parameters:

  • value_t – the data type

  • idx_t – Index type of matrix extent.

Parameters:

  • handle – [in] the raft handle.

  • truth_cluster_array – [in] the array of truth classes of type value_t

  • pred_cluster_array – [in] the array of predicted classes of type value_t

  • lower_label_range – [in] the lower bound of the range of labels

  • upper_label_range – [in] the upper bound of the range of labels

Returns:

the cluster completeness score

Cluster Dispersion#

#include <raft/stats/dispersion.cuh>

namespace raft::stats

template<typename value_t, typename idx_t>
value_t cluster_dispersion(raft::resources const &handle, raft::device_matrix_view<const value_t, idx_t, raft::row_major> centroids, raft::device_vector_view<const idx_t, idx_t> cluster_sizes, std::optional<raft::device_vector_view<value_t, idx_t>> global_centroid, const idx_t n_points)#

Compute cluster dispersion metric. This is very useful for automatically finding the ‘k’ (in kmeans) that improves this metric. The cluster dispersion metric is defined as the square root of the sum of the squared distances between the cluster centroids and the global centroid.

Template Parameters:

  • value_t – data type

  • idx_t – index type

Parameters:

  • handle – [in] the raft handle

  • centroids – [in] the cluster centroids. This is assumed to be row-major and of dimension (n_clusters x dim)

  • cluster_sizes – [in] number of points in the dataset which belong to each cluster. This is of length n_clusters

  • global_centroid – [out] compute the global weighted centroid of all cluster centroids. This is of length dim. Use std::nullopt to not return it.

  • n_points – [in] number of points in the dataset

Returns:

the cluster dispersion value

Rand Index#

#include <raft/stats/rand_index.cuh>

namespace raft::stats

template<typename value_t, typename idx_t>
double rand_index(raft::resources const &handle, raft::device_vector_view<const value_t, idx_t> first_cluster_array, raft::device_vector_view<const value_t, idx_t> second_cluster_array)#

Function to calculate RandIndex more info on rand index

Template Parameters:

  • value_t – the data type

  • idx_t – index type

Parameters:

  • handle – [in] the raft handle

  • first_cluster_array – [in] the array of classes of type value_t

  • second_cluster_array – [in] the array of classes of type value_t

Returns:

: The RandIndex value.

Silhouette Score#

#include <raft/stats/silhouette_score.cuh>

namespace raft::stats

template<typename value_t, typename label_t, typename idx_t>
value_t silhouette_score(raft::resources const &handle, raft::device_matrix_view<const value_t, idx_t, raft::row_major> X_in, raft::device_vector_view<const label_t, idx_t> labels, std::optional<raft::device_vector_view<value_t, idx_t>> silhouette_score_per_sample, idx_t n_unique_labels, raft::distance::DistanceType metric = raft::distance::DistanceType::L2Unexpanded)#

main function that returns the average silhouette score for a given set of data and its clusterings

Template Parameters:

  • value_t – type of the data samples

  • label_t – type of the labels

  • idx_t – index type

Parameters:

  • handle – [in] raft handle for managing expensive resources

  • X_in – [in] input matrix Data in row-major format (nRows x nCols)

  • labels – [in] the pointer to the array containing labels for every data sample (length: nRows)

  • silhouette_score_per_sample – [out] optional array populated with the silhouette score for every sample (length: nRows)

  • n_unique_labels – [in] number of unique labels in the labels array

  • metric – [in] the numerical value that maps to the type of distance metric to be used in the calculations

Returns:

: The silhouette score.

template<typename value_t, typename label_t, typename idx_t>
value_t silhouette_score_batched(raft::resources const &handle, raft::device_matrix_view<const value_t, idx_t, raft::row_major> X, raft::device_vector_view<const label_t, idx_t> labels, std::optional<raft::device_vector_view<value_t, idx_t>> silhouette_score_per_sample, idx_t n_unique_labels, idx_t batch_size, raft::distance::DistanceType metric = raft::distance::DistanceType::L2Unexpanded)#

function that returns the average silhouette score for a given set of data and its clusterings

Template Parameters:

  • value_t – type of the data samples

  • label_t – type of the labels

  • idx_t – index type

Parameters:

  • handle – [in] raft handle for managing expensive resources

  • X – [in] input matrix Data in row-major format (nRows x nCols)

  • labels – [in] the pointer to the array containing labels for every data sample (length: nRows)

  • silhouette_score_per_sample – [out] optional array populated with the silhouette score for every sample (length: nRows)

  • n_unique_labels – [in] number of unique labels in the labels array

  • batch_size – [in] number of samples per batch

  • metric – [in] the numerical value that maps to the type of distance metric to be used in the calculations

Returns:

: The silhouette score.

V Measure#

#include <raft/stats/v_measure.cuh>

namespace raft::stats

template<typename value_t, typename idx_t>
double v_measure(raft::resources const &handle, raft::device_vector_view<const value_t, idx_t> truth_cluster_array, raft::device_vector_view<const value_t, idx_t> pred_cluster_array, value_t lower_label_range, value_t upper_label_range, double beta = 1.0)#

Function to calculate the v-measure between two clusters.

Template Parameters:

  • value_t – the data type

  • idx_t – Integer type used to for addressing

Parameters:

  • handle – [in] the raft handle

  • truth_cluster_array – [in] the array of truth classes of type T

  • pred_cluster_array – [in] the array of predicted classes of type T

  • lower_label_range – [in] the lower bound of the range of labels

  • upper_label_range – [in] the upper bound of the range of labels

  • beta – [in] v_measure parameter

Returns:

the v-measure between the two clusters

On this page
Show Source