Struct KernelCenteringStats 

pub struct KernelCenteringStats {
    pub row_means: Array1<f64>,
    pub grand_mean: f64,
}

Training-time centering statistics kept around so that we can center the test-time kernel evaluations against the same reference frame.

Given a training kernel matrix K of shape (n, n):

• row_means[j] = (1/n) sum_i K[i, j] — the mean of column j (equivalently, row j; K is symmetric). These are the means of the training kernels with every training point.
• grand_mean = (1/n^2) sum_{i,j} K[i, j] — the double sum.

The two statistics are exactly what appears in the transformation applied to test-time kernel evaluations:

k_c(x, X_j) = k(x, X_j) - row_means[j] - (1/n) sum_i k(x, X_i) + grand_mean

Fields

row_means: Array1<f64>

Per-column means of the training Gram matrix, length n.

grand_mean: f64

Scalar grand mean of the training Gram matrix.

Implementations

impl KernelCenteringStats

pub fn n(&self) -> usize

Number of training points n.

Trait Implementations

impl Clone for KernelCenteringStats

fn clone(&self) -> KernelCenteringStats

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for KernelCenteringStats

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl PartialEq for KernelCenteringStats

fn eq(&self, other: &KernelCenteringStats) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl StructuralPartialEq for KernelCenteringStats