ml-kernel

A factory for kernel functions.

Installation

$ npm i ml-kernel

Usage

new Kernel(type, options)

This function can be called with a matrix of input vectors. and optional landmarks. If no landmark is provided, the input vectors will be used.

Available kernels:

• linear - Linear kernel
• gaussian or rbf - Gaussian (radial basis function) kernel
• polynomial or poly - Polynomial kernel
• exponential - Exponential kernel
• laplacian - Laplacian kernel
• anova - ANOVA kernel
• rational - Rational Quadratic kernel
• multiquadratic - Multiquadratic kernel
• cauchy - Cauchy kernel
• histogram or min - Histogram Intersection kernel
• sigmoid or mlp - Sigmoid (hyperbolic tangent) kernel

kernel.compute(inputs, landmarks)

This function can be called with a matrix of input vectors and optional landmarks. If no landmark is provided, the input vectors will be used.
The function returns a kernel matrix of feature space vectors.

Example

import { Kernel } from 'ml-kernel';

const kernel = new Kernel('gaussian', { sigma: 1 });
const result = kernel.compute([
  [1, 2],
  [3, 4],
]);
// result is an ml-matrix Matrix instance

License

MIT