kernels.quasisep package#

The kernels implemented in this subpackage are used with the tinygp.solvers.QuasisepSolver to allow scalable GP computations by exploting quasiseparable structure in the relevant matrices (see solvers.quasisep package for more technical details). For now, these methods are experimental, so you may find the documentation patchy in places. You are encouraged to open issues or pull requests as you find gaps.


The base class for all quasiseparable kernels


A base class for wrapping kernels with some custom implementations

Sum(kernel1, kernel2)

A helper to represent the sum of two quasiseparable kernels

Product(kernel1, kernel2)

A helper to represent the product of two quasiseparable kernels

Scale(kernel, scale)

The product of a scalar and a quasiseparable kernel

Celerite(a, b, c, d)

The baseline kernel from the celerite package

SHO(omega, quality, sigma)

The damped, driven simple harmonic oscillator kernel

Exp(scale, sigma)

A scalable implementation of tinygp.kernels.stationary.Exp

Matern32(scale, sigma)

A scalable implementation of tinygp.kernels.stationary.Matern32

Matern52(scale, sigma)

A scalable implementation of tinygp.kernels.stationary.Matern52

Cosine(scale, sigma)

A scalable implementation of tinygp.kernels.stationary.Cosine

CARMA(alpha, beta, sigma, roots, proj, ...)

A continuous autoregressive moving average (CARMA) process