pint.models.noise_model.periodic_kernel

pint.models.noise_model.periodic_kernel(nodes: ndarray, log10_sigma: float = -7, log10_ell: float = 2, log10_gam_p: float = 0, log10_p: float = 0) ndarray[source]

Quasi-periodic (SE × periodic) covariance matrix.

Matches the periodic_kernel convention in enterprise_extensions.

Parameters:
  • nodes (np.ndarray) – 1-D array of evaluation points in seconds (e.g. average TOA at each epoch).

  • log10_sigma (float) – Log10 of the amplitude, in the units of the Gaussian process coefficients (\(\mathrm{cm}^{-3}\) as used by TimeDomainSWNoise).

  • log10_ell (float) – Log10 of the squared-exponential length scale in days.

  • log10_gam_p (float) – Log10 of the periodic damping amplitude (larger -> stronger periodic decay).

  • log10_p (float) – Log10 of the periodicity in years.

Returns:

Covariance matrix of shape (len(nodes), len(nodes)).

Return type:

np.ndarray

Notes

The kernel is

\[K(\tau) = \sigma^2 \exp\!\left( -\frac{\tau^2}{2\ell^2} - \gamma_p \sin^2\!\left(\frac{\pi\tau}{p}\right) \right) + d\,\delta_{ij}\]

where \(d = (\sigma / 50000)^2\) is a small diagonal regulariser.