geopfa.extrapolation.recommend_likelihood_params

recommend_likelihood_params(Y, Y_var=None)[source]

Recommend initialization parameters and a prior for the Gaussian likelihood (observation-noise variance) used in GPy models.

Parameters:

  • Y (numpy.ndarray) – Target values of shape (N, 1) or (N,). Flattening is handled internally.

  • Y_var (float or None, optional) – Variance of the target values. If None, it is estimated using estimate_variance().

Returns:

dict –

A dictionary containing:
  • 'init'float

    Initial noise variance estimate.

  • 'prior'GPy.priors.Gamma

    Gamma prior placed on the noise variance.

  • 'bounds'tuple(float, float)

    Lower and upper bounds for the noise variance parameter.

The dictionary is packaged under the key 'variance' so that callers can directly configure model.Gaussian_noise.variance.

Notes

  • The initial noise variance is set to approximately 1% of the variance of Y.

  • A Gamma prior is used to bias the optimization toward small but positive values, reflecting typical assumptions about measurement noise.