Preprocessing PCA

Source header: cuvs/preprocessing/pca.hpp

Types

preprocessing::pca::params

Parameters for PCA decomposition. Ref:

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

1 struct params { ... };

Fields

Name	Type	Description
`n_components`	`int`	Number of components to keep.
`copy`	`bool`	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.
`whiten`	`bool`	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.
`algorithm`	`solver`	The solver algorithm to use.
`tol`	`float`	Tolerance for singular values computed by svd_solver == ‘arpack’ or the Jacobi solver.
`n_iterations`	`int`	Number of iterations for the power method computed by the Jacobi solver.

PCA (Principal Component Analysis)

preprocessing::pca::fit

Perform PCA fit operation.

1 void fit(raft::resources const& handle,
2 const params& config,
3 raft::device_matrix_view<float, int64_t, raft::col_major> input,
4 raft::device_matrix_view<float, int64_t, raft::col_major> components,
5 raft::device_vector_view<float, int64_t> explained_var,
6 raft::device_vector_view<float, int64_t> explained_var_ratio,
7 raft::device_vector_view<float, int64_t> singular_vals,
8 raft::device_vector_view<float, int64_t> mu,
9 raft::device_scalar_view<float, int64_t> noise_vars,
10 bool flip_signs_based_on_U = false);

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

Parameters

Name	Direction	Type	Description
`handle`	in	`raft::resources const&`	raft resource handle
`config`	in	`const params&`	PCA parameters
`input`	inout	`raft::device_matrix_view<float, int64_t, raft::col_major>`	input data [n_rows x n_cols] (col-major). Modified temporarily.
`components`	out	`raft::device_matrix_view<float, int64_t, raft::col_major>`	principal components [n_components x n_cols] (col-major)
`explained_var`	out	`raft::device_vector_view<float, int64_t>`	explained variances [n_components]
`explained_var_ratio`	out	`raft::device_vector_view<float, int64_t>`	explained variance ratios [n_components]
`singular_vals`	out	`raft::device_vector_view<float, int64_t>`	singular values [n_components]
`mu`	out	`raft::device_vector_view<float, int64_t>`	column means [n_cols]
`noise_vars`	out	`raft::device_scalar_view<float, int64_t>`	noise variance (scalar)
`flip_signs_based_on_U`	in	`bool`	whether to determine signs by U (true) or V.T (false) Default: `false`.

Returns

void

preprocessing::pca::fit_transform

Perform PCA fit and transform operations.

1 void fit_transform(raft::resources const& handle,
2 const params& config,
3 raft::device_matrix_view<float, int64_t, raft::col_major> input,
4 raft::device_matrix_view<float, int64_t, raft::col_major> trans_input,
5 raft::device_matrix_view<float, int64_t, raft::col_major> components,
6 raft::device_vector_view<float, int64_t> explained_var,
7 raft::device_vector_view<float, int64_t> explained_var_ratio,
8 raft::device_vector_view<float, int64_t> singular_vals,
9 raft::device_vector_view<float, int64_t> mu,
10 raft::device_scalar_view<float, int64_t> noise_vars,
11 bool flip_signs_based_on_U = false);

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

Parameters

Name	Direction	Type	Description
`handle`	in	`raft::resources const&`	raft resource handle
`config`	in	`const params&`	PCA parameters
`input`	inout	`raft::device_matrix_view<float, int64_t, raft::col_major>`	input data [n_rows x n_cols] (col-major). Modified temporarily.
`trans_input`	out	`raft::device_matrix_view<float, int64_t, raft::col_major>`	transformed data [n_rows x n_components] (col-major)
`components`	out	`raft::device_matrix_view<float, int64_t, raft::col_major>`	principal components [n_components x n_cols] (col-major)
`explained_var`	out	`raft::device_vector_view<float, int64_t>`	explained variances [n_components]
`explained_var_ratio`	out	`raft::device_vector_view<float, int64_t>`	explained variance ratios [n_components]
`singular_vals`	out	`raft::device_vector_view<float, int64_t>`	singular values [n_components]
`mu`	out	`raft::device_vector_view<float, int64_t>`	column means [n_cols]
`noise_vars`	out	`raft::device_scalar_view<float, int64_t>`	noise variance (scalar)
`flip_signs_based_on_U`	in	`bool`	whether to determine signs by U (true) or V.T (false) Default: `false`.

Returns

void

preprocessing::pca::transform

Perform PCA transform operation.

1 void transform(raft::resources const& handle,
2 const params& config,
3 raft::device_matrix_view<float, int64_t, raft::col_major> input,
4 raft::device_matrix_view<float, int64_t, raft::col_major> components,
5 raft::device_vector_view<float, int64_t> singular_vals,
6 raft::device_vector_view<float, int64_t> mu,
7 raft::device_matrix_view<float, int64_t, raft::col_major> trans_input);

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

(mean-centered then restored).

Parameters

Name	Direction	Type	Description
`handle`	in	`raft::resources const&`	raft resource handle
`config`	in	`const params&`	PCA parameters
`input`	inout	`raft::device_matrix_view<float, int64_t, raft::col_major>`	data to transform [n_rows x n_cols] (col-major). Modified temporarily
`components`	in	`raft::device_matrix_view<float, int64_t, raft::col_major>`	principal components [n_components x n_cols] (col-major)
`singular_vals`	in	`raft::device_vector_view<float, int64_t>`	singular values [n_components]
`mu`	in	`raft::device_vector_view<float, int64_t>`	column means [n_cols]
`trans_input`	out	`raft::device_matrix_view<float, int64_t, raft::col_major>`	transformed data [n_rows x n_components] (col-major)

Returns

void

preprocessing::pca::inverse_transform

Perform PCA inverse transform operation.

1 void inverse_transform(raft::resources const& handle,
2 const params& config,
3 raft::device_matrix_view<float, int64_t, raft::col_major> trans_input,
4 raft::device_matrix_view<float, int64_t, raft::col_major> components,
5 raft::device_vector_view<float, int64_t> singular_vals,
6 raft::device_vector_view<float, int64_t> mu,
7 raft::device_matrix_view<float, int64_t, raft::col_major> output);

Transforms data from the eigenspace back to the original space.

Parameters

Name	Direction	Type	Description
`handle`	in	`raft::resources const&`	raft resource handle
`config`	in	`const params&`	PCA parameters
`trans_input`	in	`raft::device_matrix_view<float, int64_t, raft::col_major>`	transformed data [n_rows x n_components] (col-major)
`components`	in	`raft::device_matrix_view<float, int64_t, raft::col_major>`	principal components [n_components x n_cols] (col-major)
`singular_vals`	in	`raft::device_vector_view<float, int64_t>`	singular values [n_components]
`mu`	in	`raft::device_vector_view<float, int64_t>`	column means [n_cols]
`output`	out	`raft::device_matrix_view<float, int64_t, raft::col_major>`	reconstructed data [n_rows x n_cols] (col-major)

Returns

void