PiecewisePolynomialKernel<T>

Models & Types · AiDotNet.Kernels

Implements the Piecewise Polynomial kernel for measuring similarity between data points.

Implements: IKernelFunction<T>

For Beginners

A kernel function is a mathematical tool that measures how similar two data points are. The Piecewise Polynomial kernel is special because it has a “cutoff distance” - if two points are farther apart than this distance, the kernel says they have zero similarity.

How It Works

The Piecewise Polynomial kernel is a compact kernel function that produces a similarity measure that becomes exactly zero when points are far enough apart.

This property makes the Piecewise Polynomial kernel useful for:

  • Speeding up calculations in large datasets (since many calculations become zero)
  • Problems where you only want nearby points to influence each other
  • Creating sparse matrices in kernel methods

The formula for this kernel is: k(x, y) = (1 - ||x-y||/c)^(j+1) if ||x-y|| = c, and 0 otherwise where:

  • x and y are the two data points being compared
  • ||x-y|| is the Euclidean distance between them
  • c is the cutoff distance parameter
  • j is the degree parameter

Example

using AiDotNet.Kernels;
using AiDotNet.Tensors.LinearAlgebra;

var kernel = new PiecewisePolynomialKernel<double>();
var a = new Vector<double>(new[] { 1.0, 2.0, 3.0 });
var b = new Vector<double>(new[] { 1.5, 1.0, 2.5 });
double similarity = kernel.Calculate(a, b);
Console.WriteLine($"PiecewisePolynomialKernel: K(a, b) = {similarity:F4}");

Constructors

Constructor Summary
PiecewisePolynomialKernel(int degree, T c) Initializes a new instance of the Piecewise Polynomial kernel with optional degree and cutoff parameters.

Methods

Method Summary
T Calculate(Vector<T> x1, Vector<T> x2) Calculates the Piecewise Polynomial kernel value between two vectors.