This is an implementation of Lomb-Scargle periodogram with several non-conventional benefits:

• Roboust false positive probability computations (data can be non-Gaussian with almost arbitrary systematics). This is achieved by constructing special null signal templates (NST) that act as effective null simulations when used on the real data.
• Periodogram score takes into account correlated noise.
• Typically, periodogram score is the likelihood ratio, here it can also be the Bayes factor, marginalizing over the parameters of the noise correlation kernel and the frequency of the signal.
• Everything is implemented in JAX with all the usual benefits: easy parallelization, GPU support, speed, etc.

Example Application: Reanalysis of Supermassive Black Hole Binaries search from here is on the 'quasars' branch.

To get started, checkout the tutorial. See also the associated paper.

If you encounter any issues, feel free to contact me at [email protected] .