Module ard_kernel 

Automatic Relevance Determination (ARD) kernels.

ARD kernels learn a separate length scale for each input dimension, automatically determining the relevance of each feature. This is particularly useful for high-dimensional data where some features may be more important than others.

Key Features

• Per-dimension length scales: Each feature has its own length scale
• Feature selection: Irrelevant features get large length scales (effectively ignored)
• Gradient support: For hyperparameter optimization via gradient descent

Example

use tensorlogic_sklears_kernels::ard_kernel::{ArdRbfKernel, ArdMaternKernel};
use tensorlogic_sklears_kernels::Kernel;

// Create ARD RBF kernel with 3 features, each with its own length scale
let length_scales = vec![1.0, 2.0, 0.5]; // Different relevance per dimension
let kernel = ArdRbfKernel::new(length_scales.clone()).unwrap();

let x = vec![1.0, 2.0, 3.0];
let y = vec![1.5, 2.5, 3.5];
let sim = kernel.compute(&x, &y).unwrap();

Structs

ArdMaternKernel

ARD Matérn kernel with per-dimension length scales.

ArdRationalQuadraticKernel

ARD Rational Quadratic kernel.

ArdRbfKernel

ARD (Automatic Relevance Determination) RBF kernel.

ConstantKernel

Constant kernel: K(x, y) = σ²

DotProductKernel

Dot Product kernel (Linear kernel with variance and shift).

KernelGradient

Gradient information for kernel hyperparameter optimization.

ScaledKernel

Scaled kernel wrapper that multiplies a kernel by a variance parameter.

WhiteNoiseKernel

Utility kernel: White Noise kernel for observation noise modeling.