TanimotoKernel<T>

Models & Types · AiDotNet.Kernels

Implements the Tanimoto kernel (also known as the Jaccard 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 Tanimoto kernel is special because it focuses on the overlap between features relative to their combined magnitude. It’s like measuring how much two sets have in common compared to their total size.

How It Works

The Tanimoto kernel is a similarity measure that is particularly useful for binary data and chemical fingerprints, but can be applied to any vector data. It measures the ratio of the intersection to the union of the features.

Think of it like this: Imagine you have two shopping lists. The Tanimoto kernel would measure how many items appear on both lists (the overlap) divided by the total number of unique items across both lists. A value of 1 means the lists are identical, while a value close to 0 means they have very little in common.

The formula for the Tanimoto kernel is: k(x, y) = (x · y) / (||x||² + ||y||² - x · y) where:

  • x · y is the dot product between vectors x and y
  • ||x||² is the squared norm of vector x (dot product of x with itself)
  • ||y||² is the squared norm of vector y (dot product of y with itself)

Common uses include:

  • Comparing chemical compounds (molecular fingerprints)
  • Document similarity in text analysis
  • Binary feature vectors in machine learning
  • Recommendation systems

Example

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

var kernel = new TanimotoKernel<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($"TanimotoKernel: K(a, b) = {similarity:F4}");

Constructors

Constructor Summary
TanimotoKernel() Initializes a new instance of the Tanimoto kernel.

Methods

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