Spectral

Source header: cuvs/cluster/spectral.hpp

Spectral Clustering Parameters

cluster::spectral::params

Parameters for spectral clustering

1 struct params {
2   int n_clusters;
3   int n_components;
4   int n_init;
5   int n_neighbors;
6   float tolerance;
7   raft::random::RngState rng_state;
8 };

Fields

Name	Type	Description
`n_clusters`	`int`	Number of clusters to find
`n_components`	`int`	Number of eigenvectors to use for the spectral embedding (typically equal to n_clusters)
`n_init`	`int`	Number of k-means runs with different centroid seeds
`n_neighbors`	`int`	Number of nearest neighbors for constructing the connectivity graph
`tolerance`	`float`	Tolerance for the eigenvalue solver
`rng_state`	`raft::random::RngState`	Random number generator state for reproducibility

Spectral Clustering

cluster::spectral::fit_predict

Perform spectral clustering on a connectivity graph

1 void fit_predict(raft::resources const& handle,
2 params config,
3 raft::device_coo_matrix_view<float, int, int, int> connectivity_graph,
4 raft::device_vector_view<int, int> labels);

Parameters

Name	Direction	Type	Description
`handle`	in	`raft::resources const&`	RAFT resource handle
`config`	in	`params`	Spectral clustering parameters
`connectivity_graph`	in	`raft::device_coo_matrix_view<float, int, int, int>`	Sparse COO matrix representing connectivity between data points
`labels`	out	`raft::device_vector_view<int, int>`	Device vector of size n_samples to store cluster assignments (0 to n_clusters-1)

Returns

void

Additional overload: cluster::spectral::fit_predict

Perform spectral clustering on a connectivity graph

1 void fit_predict(raft::resources const& handle,
2 params config,
3 raft::device_coo_matrix_view<double, int, int, int> connectivity_graph,
4 raft::device_vector_view<int, int> labels);

Parameters

Name	Direction	Type	Description
`handle`	in	`raft::resources const&`	RAFT resource handle
`config`	in	`params`	Spectral clustering parameters
`connectivity_graph`	in	`raft::device_coo_matrix_view<double, int, int, int>`	Sparse COO matrix representing connectivity between data points
`labels`	out	`raft::device_vector_view<int, int>`	Device vector of size n_samples to store cluster assignments (0 to n_clusters-1)

Returns

void

Additional overload: cluster::spectral::fit_predict

Perform spectral clustering on a dense dataset

1 void fit_predict(raft::resources const& handle,
2 params config,
3 raft::device_matrix_view<float, int, raft::row_major> dataset,
4 raft::device_vector_view<int, int> labels);

This overload automatically constructs the connectivity graph from the input dataset using k-nearest neighbors.

Parameters

Name	Direction	Type	Description
`handle`	in	`raft::resources const&`	RAFT resource handle
`config`	in	`params`	Spectral clustering parameters
`dataset`	in	`raft::device_matrix_view<float, int, raft::row_major>`	Dense row-major matrix of shape (n_samples, n_features)
`labels`	out	`raft::device_vector_view<int, int>`	Device vector of size n_samples to store cluster assignments (0 to n_clusters-1)

Returns

void

1	struct params {
2	int n_clusters;
3	int n_components;
4	int n_init;
5	int n_neighbors;
6	float tolerance;
7	raft::random::RngState rng_state;
8	};

1	void fit_predict(raft::resources const& handle,
2	params config,
3	raft::device_coo_matrix_view<float, int, int, int> connectivity_graph,
4	raft::device_vector_view<int, int> labels);

1	void fit_predict(raft::resources const& handle,
2	params config,
3	raft::device_coo_matrix_view<double, int, int, int> connectivity_graph,
4	raft::device_vector_view<int, int> labels);

1	void fit_predict(raft::resources const& handle,
2	params config,
3	raft::device_matrix_view<float, int, raft::row_major> dataset,
4	raft::device_vector_view<int, int> labels);