Ctrl+K

comet_maths.interpolation.interpolation.gaussian_process_regression

comet_maths.interpolation.interpolation.gaussian_process_regression#

comet_maths.interpolation.interpolation.gaussian_process_regression(x_i: ndarray, y_i: ndarray, x: ndarray, u_y_i: ndarray | None = None, corr_y_i: ndarray | str | None = None, kernel: str | None = 'RBF', min_scale: float | None = 0.01, max_scale: float | None = 10000, extrapolate: str | None = 'extrapolate', return_uncertainties: bool | None = True, return_corr: bool | None = False, include_model_uncertainties: bool | None = True, add_model_error: bool | None = False, MCsteps: int | None = 100, parallel_cores: int | None = 4) → ndarray[source]#

Function to perform interpolation using Gaussian process regression

Parameters:

  • x_i – Independent variable quantity x (coordinate data of y_i)

  • y_i – measured variable quantity y (data to interpolate)

  • x – Independent variable quantity x for which we are trying to obtain the measurand y

  • u_y_i – uncertainties on y_i, defaults to None

  • corr_y_i – error correlation matrix (can be “rand” for random, “syst” for systematic, or a custom 2D error correlation matrix), defaults to None

  • kernel – kernel to be used in the gpr interpolation. Defaults to “RBF”.

  • min_scale – minimum bound on the scale parameter in the gaussian process regression. Defaults to 0.01

  • max_scale – maximum bound on the scale parameter in the gaussian process regression. Defaults to 100

  • extrapolate – extrapolation method, which can be set to “extrapolate” (in which case extrapolation is used using interpolation method defined in “method”), “nearest” (in which case nearest values are used for extrapolation), or “linear” (in which case linear extrapolation is used). Defaults to “extrapolate”.

  • return_uncertainties – Boolean to indicate whether interpolation uncertainties should be calculated and returned. Defaults to False

  • return_corr – Boolean to indicate whether interpolation error-correlation matrix should be calculated and returned. Defaults to False

  • include_model_uncertainties – Boolean to indicate whether model uncertainties should be added to output uncertainties to account for interpolation uncertainties. Not used for gpr. Defaults to True

  • add_model_error – Boolean to indicate whether model error should be added to interpolated values to account for interpolation errors (useful in Monte Carlo approaches). Defaults to False

  • MCsteps – number of MC iterations. Defaults to 100

  • parallel_cores – number of CPU to be used in parallel processing. Defaults to 4

Returns:

The measurand y evaluated at the values x (interpolated data)

Contents