noise package

noise package#

This subpackage provides the tools needed to build expressive observation processes for tinygp Gaussian process models. The most commonly used noise model is Diagonal, which adds a constant diagonal matrix to the process covariance to represent per-observation noise. This subpackage also includes a Dense model for adding a full rank observation model, and Banded to capture noise that can be represented by a banded matrix.


A diagonal observation noise model


A full rank observation noise model

Banded(diag, off_diags)

A banded observation noise model