Kernels#

RBF

The Radial Basis Function (RBF) kernel.

RFF

Computes an approximation of the kernel using Random Fourier Features.

AbstractKernel

Base kernel class.

ArcCosine

The ArCosine kernel.

BasisFunctionComputation

Compute engine class for finite basis function approximations to a kernel.

Constant

A constant kernel.

ConstantDiagonalKernelComputation

Computation engine for constant diagonal kernels.

DenseKernelComputation

Dense kernel computation class.

DiagonalKernelComputation

Diagonal kernel computation class.

EigenKernelComputation

Eigen kernel computation class.

Gibbs

A base kernel whose lengthscale changes with location.

Gneiting

The Gneiting nonseparable space–time kernel.

GraphKernel

The Matérn graph kernel defined on the vertex set of a graph.

ICMKernel

Intrinsic Coregionalization Model kernel.

LCMKernel

Linear Model of Coregionalization kernel.

Linear

The linear kernel.

Matern12

The Matérn kernel with smoothness parameter fixed at 0.5.

Matern32

The Matérn kernel with smoothness parameter fixed at 1.5.

Matern52

The Matérn kernel with smoothness parameter fixed at 2.5.

MultiOutputKernel

Base class for multi-output kernels.

MultiOutputKernelComputation

Compute engine for multi-output kernels.

OrthogonalAdditiveKernel

Orthogonal Additive Kernel (OAK).

Periodic

The periodic kernel.

Polynomial

The Polynomial kernel with variable degree.

PoweredExponential

The powered exponential family of kernels.

ProductKernel

A kernel that is the product of a set of kernels.

RationalQuadratic

The Rational Quadratic kernel.

SumKernel

A kernel that is the sum of a set of kernels.

VaryingAmplitude

A base kernel whose standard deviation changes with location.

White

The White noise kernel.

Location functions#

A location function gives a kernel parameter that changes with input location, such as the standard deviation of VaryingAmplitude or the lengthscale of Gibbs. It is not a mean function: a mean function describes the Gaussian process, but a location function describes its covariance. A location function evaluates one input point, selects its own columns with active_dims, and returns a value on the log scale.

AbstractLocationFunction

Base class for location functions.

Constant

A location function with the same value at all locations.

Linear

A location function that is log-linear in selected input columns.