Module friedrich::kernel

Kernels

A kernel is a function that maps from two row vectors to a scalar which is used to express the similarity between the vectors.

To learn more about the properties of the provided kernels, we recommand the Usual_covariance_functions Wikipedia page and the kernel-functions-for-machine-learning-applications article.

User-defined kernels should implement the Kernel trait. To learn more about the implementation of kernels adapted to a particular problem, we recommend the chapter two (Expressing Structure with Kernels) and three (Automatic Model Construction) of the very good Automatic Model Construction with Gaussian Processes.

This implementation is inspired by rusty-machines'.

Structs

Exponential	The Exponential Kernel
HyperTan	The Hyperbolic Tangent Kernel.
KernelArith	A wrapper tuple struct used for kernel arithmetic
KernelProd	The pointwise product of two kernels
KernelSum	The sum of two kernels
Linear	The Linear Kernel
Matern1	The Matèrn1 kernel which is 1 differentiable and correspond to a classical Matèrn kernel with nu=3/2
Matern2	The Matèrn2 kernel which is 2 differentiable and correspond to a classical Matèrn kernel with nu=5/2
Multiquadric	The Multiquadric Kernel.
Polynomial	The Polynomial Kernel
RationalQuadratic	The Rational Quadratic Kernel.
SquaredExp	Squared exponential kernel

Traits

Kernel

The Kernel trait

Type Definitions

Gaussian

Gaussian kernel