Struct GlobalAlignmentKernel 

pub struct GlobalAlignmentKernel {
    pub sigma: f64,
    pub gamma: f64,
    pub normalize: bool,
}

Global Alignment Kernel (GAK) for time series

GAK is a kernel that computes similarity between time series using a global alignment approach with a triangular smoothing kernel. It’s related to DTW but uses a differentiable soft-DTW formulation.

K_GAK(x, y, σ) = exp(-γ * GA_σ(x, y))

where GA_σ is the global alignment distance with bandwidth σ.

References:

• Cuturi, M. (2011). Fast global alignment kernels.
• Cuturi, M. & Blondel, M. (2017). Soft-DTW: a differentiable loss function for time-series.

Fields

sigma: f64

Bandwidth parameter for triangular kernel

gamma: f64

Gamma parameter for RBF transformation

normalize: bool

Whether to normalize

Implementations

impl GlobalAlignmentKernel

pub fn new(sigma: f64, gamma: f64) -> Self

Create a new GAK kernel

pub fn compute_gak_similarity( &self, seq1: &Array1<f64>, seq2: &Array1<f64>, ) -> f64

Compute GAK similarity between two sequences

Trait Implementations

impl Clone for GlobalAlignmentKernel

fn clone(&self) -> GlobalAlignmentKernel

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for GlobalAlignmentKernel

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

Formats the value using the given formatter.

impl Default for GlobalAlignmentKernel

fn default() -> Self

Returns the “default value” for a type.

impl Kernel for GlobalAlignmentKernel

fn compute(&self, x: ArrayView1<'_, f64>, y: ArrayView1<'_, f64>) -> f64

Compute kernel value between two vectors

fn parameters(&self) -> HashMap<String, f64>

Get kernel parameters for serialization

fn compute_matrix(&self, x: &Array2<f64>, y: &Array2<f64>) -> Array2<f64>

Compute kernel matrix for two datasets