Telerite

Telerite is a Python software library that implements fast Gaussian process regression for time-series data, with particular emphasis on astronomical applications such as stellar photometry and radial-velocity measurements. It provides a practical framework for modeling correlated noise and signals in unevenly sampled data.

The core of telerite is the celerite algorithm, a method that structures the covariance of the Gaussian process in a way that allows linear-time training and evaluation for certain kernel families. This dramatically reduces the computational burden compared with generic Gaussian processes, enabling analysis of larger time-series datasets.

Kernels in telerite are built from simple, interpretable components, notably exponentials and damped harmonic-oscillator terms, which can be combined to model quasi-periodic, stochastic, or slowly varying signals typical of astrophysical time series. This flexibility supports modeling of stellar activity, instrumental systematics, and other sources of correlated noise alongside potential signals of interest such as exoplanet transits or oscillations.

The library provides a Python interface built on NumPy and SciPy, supports irregular sampling and heteroscedastic measurement uncertainties, and can be used with standard optimization or sampling tools to infer hyperparameters or posterior distributions. It integrates with common scientific workflows and is designed for numerical stability in large-scale analyses.

In practical use, telerite is employed to whiten or model correlated noise in light curves and radial-velocity data, to separate planetary or transit signals from stellar activity, and to perform rapid exploratory analyses on extensive datasets. The project is open source, with documentation and tutorials available to guide installation and application.