Struct GpuPCA 

pub struct GpuPCA {
    pub n_components: usize,
    pub center: bool,
    pub components: Option<Array2<f64>>,
    pub explained_variance: Option<Array1<f64>>,
    pub mean: Option<Array1<f64>>,
    /* private fields */
}

GPU-accelerated Principal Component Analysis

Fields

n_components: usize

Number of components to compute

center: bool

Whether to center the data

components: Option<Array2<f64>>

Principal components (loading vectors)

explained_variance: Option<Array1<f64>>

Explained variance for each component

mean: Option<Array1<f64>>

Mean values for centering

Implementations

impl GpuPCA

pub fn new(n_components: usize) -> Result<Self>

Create a new GPU PCA instance

Arguments

• n_components - Number of principal components to compute

Returns

Returns a new GpuPCA instance with GPU context initialized

Errors

Returns an error if GPU initialization fails or if n_components is 0

Examples

let pca = GpuPCA::new(5)?;

pub fn fit(&mut self, x: &ArrayView2<'_, f64>) -> Result<()>

Fit the PCA model using a CPU SVD-based backend.

Currently delegates to the CPU SVD-based PCA implementation. A wgpu-backed Jacobi SVD path is planned for v0.5.

Arguments

• x - Input data matrix with shape (n_samples, n_features)

Errors

Returns an error if the input is invalid or if n_components exceeds min(n_samples, n_features).

Examples

let mut pca = GpuPCA::new(2)?;
let data = Array2::from_shape_vec((5, 4), vec![
    1.0, 2.0, 3.0, 4.0,
    2.0, 3.0, 4.0, 5.0,
    3.0, 4.0, 5.0, 6.0,
    4.0, 5.0, 6.0, 7.0,
    5.0, 6.0, 7.0, 8.0,
])?;
pca.fit(&data.view())?;
assert!(pca.components.is_some());

pub fn transform(&self, x: &ArrayView2<'_, f64>) -> Result<Array2<f64>>

Transform data using the fitted PCA model.

Projects the input data onto the principal components computed during GpuPCA::fit. Currently runs on the CPU; a wgpu-backed path is planned for v0.5.

Arguments

• x - Input data matrix with shape (n_samples, n_features)

Returns

Transformed data matrix with shape (n_samples, n_components)

Errors

Returns NotFitted if GpuPCA::fit has not been called yet, or InvalidInput if the number of features does not match.

pub fn fit_transform(&mut self, x: &ArrayView2<'_, f64>) -> Result<Array2<f64>>

Fit the PCA model and project data in one step.

Equivalent to calling GpuPCA::fit followed by GpuPCA::transform on the same data.

Arguments

• x - Input data matrix with shape (n_samples, n_features)

Returns

Transformed data matrix with shape (n_samples, n_components)

Errors

Returns any error produced by GpuPCA::fit or GpuPCA::transform.

pub fn explained_variance_ratio(&self) -> Result<Array1<f64>>

Get the explained variance ratio for each principal component.

The returned array sums to approximately 1.0 after fitting.

Returns

Array of explained variance ratios with length n_components

Errors

Returns NotFitted if GpuPCA::fit has not been called yet.