covariance_kernels

Covariance kernels for Gaussian processes.

Useful for discrepancy modeling. See pmrf.stats.discrepancy_models for more details.

Classes

AbstractCovarianceKernel(*[, name, metadata])	Abstract base class for covariance kernel functions.
AutoCrossKernel(auto, cross, num_outputs, *)	Kernel that routes between a auto-correlation and cross-correlation kernels.
ConstantKernel(variance, *[, name, metadata])	Kernel that returns a constant variance.
CosineKernel(period, *[, name, metadata])	Cosine kernel.
Matern32Kernel(lengthscale, *[, name, metadata])	Matérn kernel with nu=3/2.
Matern52Kernel(lengthscale, *[, name, metadata])	Matérn kernel with nu=5/2.
PeriodicKernel(period, lengthscale, *[, ...])	Periodic (Exp-Sine-Squared) kernel.
ProductKernel(k1, k2, *[, name, metadata])	Kernel representing the product of two kernels.
RBFKernel(lengthscale, *[, name, metadata])	Radial Basis Function (Squared Exponential) kernel.
SharedIndependentKernel(base_kernel[, ...])	Shares a base kernel across trailing independent batch axes (e.g. the real and imaginary parts), so that every element on those axes uses the same hyperparameters.
SumKernel(k1, k2, *[, name, metadata])	Kernel representing the sum of two kernels.
WhiteNoiseKernel(variance, *[, name, metadata])	Kernel representing independent Gaussian noise.
ZeroKernel(*[, name, metadata])	Kernel that always evaluates to zero.

Functions

cross_gram(kernel, x1, x2)	Build the cross-Gram matrix of a kernel between two sets of inputs.
gram(kernel, x, *[, jitter])	Build the Gram (covariance) matrix of a kernel evaluated at a set of inputs.