PCA#

Principal Component Analysis (PCA) is a linear dimensionality reduction technique that projects data onto orthogonal directions of maximum variance.

#include <cuvs/preprocessing/pca.hpp>

namespace cuvs::preprocessing::pca

Params#

struct params#

Parameters for PCA decomposition. Ref: http://scikit-learn.org/stable/modules/generated/sklearn.decomposition.PCA.html.

Public Members

int n_components = 1#: Number of components to keep.

bool copy = true#: If false, data passed to fit are overwritten and running fit(X).transform(X) will not yield the expected results, use fit_transform(X) instead.

bool whiten = false#: When true (false by default) the components vectors are multiplied by the square root of n_samples and then divided by the singular values to ensure uncorrelated outputs with unit component-wise variances.

solver algorithm = solver::COV_EIG_DQ#: The solver algorithm to use.

float tol = 0.0f#: Tolerance for singular values computed by svd_solver == ‘arpack’ or the Jacobi solver.

int n_iterations = 15#: Number of iterations for the power method computed by the Jacobi solver.

PCA#

void fit( raft::resources const &handle, const params &config, raft::device_matrix_view<float, int64_t, raft::col_major> input, raft::device_matrix_view<float, int64_t, raft::col_major> components, raft::device_vector_view<float, int64_t> explained_var, raft::device_vector_view<float, int64_t> explained_var_ratio, raft::device_vector_view<float, int64_t> singular_vals, raft::device_vector_view<float, int64_t> mu, raft::device_scalar_view<float, int64_t> noise_vars, bool flip_signs_based_on_U = false )#

Perform PCA fit operation.

Computes the principal components, explained variances, singular values, and column means from the input data.

#include <raft/core/resources.hpp>
#include <cuvs/preprocessing/pca.hpp>

raft::resources handle;

cuvs::preprocessing::pca::params params;
params.n_components = 2;

auto input = raft::make_device_matrix<float, int>(handle, n_rows, n_cols);
// ... fill input ...

auto components       = raft::make_device_matrix<float, int, raft::col_major>(
    handle, params.n_components, n_cols);
auto explained_var    = raft::make_device_vector<float, int>(handle, params.n_components);
auto explained_var_ratio = raft::make_device_vector<float, int>(handle, params.n_components);
auto singular_vals    = raft::make_device_vector<float, int>(handle, params.n_components);
auto mu               = raft::make_device_vector<float, int>(handle, n_cols);
auto noise_vars       = raft::make_device_scalar<float>(handle);

cuvs::preprocessing::pca::fit(handle, params,
    input.view(), components.view(), explained_var.view(),
    explained_var_ratio.view(), singular_vals.view(), mu.view(), noise_vars.view());

Parameters:

handle – [in] raft resource handle
config – [in] PCA parameters
input – [inout] input data [n_rows x n_cols] (col-major). Modified temporarily.
components – [out] principal components [n_components x n_cols] (col-major)
explained_var – [out] explained variances [n_components]
explained_var_ratio – [out] explained variance ratios [n_components]
singular_vals – [out] singular values [n_components]
mu – [out] column means [n_cols]
noise_vars – [out] noise variance (scalar)
flip_signs_based_on_U – [in] whether to determine signs by U (true) or V.T (false)

void fit_transform( raft::resources const &handle, const params &config, raft::device_matrix_view<float, int64_t, raft::col_major> input, raft::device_matrix_view<float, int64_t, raft::col_major> trans_input, raft::device_matrix_view<float, int64_t, raft::col_major> components, raft::device_vector_view<float, int64_t> explained_var, raft::device_vector_view<float, int64_t> explained_var_ratio, raft::device_vector_view<float, int64_t> singular_vals, raft::device_vector_view<float, int64_t> mu, raft::device_scalar_view<float, int64_t> noise_vars, bool flip_signs_based_on_U = false )#

Perform PCA fit and transform operations.

Computes the principal components and transforms the input data into the eigenspace in a single operation.

Parameters:

handle – [in] raft resource handle
config – [in] PCA parameters
input – [inout] input data [n_rows x n_cols] (col-major). Modified temporarily.
trans_input – [out] transformed data [n_rows x n_components] (col-major)
components – [out] principal components [n_components x n_cols] (col-major)
explained_var – [out] explained variances [n_components]
explained_var_ratio – [out] explained variance ratios [n_components]
singular_vals – [out] singular values [n_components]
mu – [out] column means [n_cols]
noise_vars – [out] noise variance (scalar)
flip_signs_based_on_U – [in] whether to determine signs by U (true) or V.T (false)

void transform( raft::resources const &handle, const params &config, raft::device_matrix_view<float, int64_t, raft::col_major> input, raft::device_matrix_view<float, int64_t, raft::col_major> components, raft::device_vector_view<float, int64_t> singular_vals, raft::device_vector_view<float, int64_t> mu, raft::device_matrix_view<float, int64_t, raft::col_major> trans_input )#

Perform PCA transform operation.

Transforms the input data into the eigenspace using previously computed principal components.

Parameters:

handle – [in] raft resource handle
config – [in] PCA parameters
input – [inout] data to transform [n_rows x n_cols] (col-major). Modified temporarily (mean-centered then restored).
components – [in] principal components [n_components x n_cols] (col-major)
singular_vals – [in] singular values [n_components]
mu – [in] column means [n_cols]
trans_input – [out] transformed data [n_rows x n_components] (col-major)

void inverse_transform( raft::resources const &handle, const params &config, raft::device_matrix_view<float, int64_t, raft::col_major> trans_input, raft::device_matrix_view<float, int64_t, raft::col_major> components, raft::device_vector_view<float, int64_t> singular_vals, raft::device_vector_view<float, int64_t> mu, raft::device_matrix_view<float, int64_t, raft::col_major> output )#

Perform PCA inverse transform operation.

Transforms data from the eigenspace back to the original space.

Parameters:

handle – [in] raft resource handle
config – [in] PCA parameters
trans_input – [in] transformed data [n_rows x n_components] (col-major)
components – [in] principal components [n_components x n_cols] (col-major)
singular_vals – [in] singular values [n_components]
mu – [in] column means [n_cols]
output – [out] reconstructed data [n_rows x n_cols] (col-major)