nifty.re.multi_grid.correlated_field module#

class ICRField(grid: nifty.re.multi_grid.grid.Grid, kernel: dict | nifty.re.model.Model | Callable[[NDArray, NDArray], NDArray], *, offset=0.0, window_size=None, compress: bool | dict = {'rtol': 1e-05, 'atol': 1e-10, 'buffer_size': 10000, 'use_distances': True}, prefix='mgcfm')[source]#

Bases: Model

__init__(grid: Grid, kernel: dict | Model | Callable[[NDArray, NDArray], NDArray], *, offset=0.0, window_size=None, compress: bool | dict = {'atol': 1e-10, 'buffer_size': 10000, 'rtol': 1e-05, 'use_distances': True}, prefix='mgcfm')[source]#

Correlated field model utilizing ICR to work on arbitrarily charted grids.

Parameters:

  • grid (Grid) – Storage object with instructions on how pixels are spaced.

  • kernel (Model or callable or dict) –

    One of the following: - NIFTy Model which yields a callable covariance function (kernel(xi: dict) -> callable[[x, y], z]) - callable[[x, y], z] covariance function - dict with the parameters of the experimental Matern kernel (deprecated)

    Parameters for the dict option: - kind: str – Type of the kernel. Only “experimental_matern” is supported. - scale: tuple or Model or float – Prior scale of the kernel. - cutoff: tuple or Model or float – Prior cutoff of the covariance kernel. - loglogslope: tuple or Model or float – Prior power-law slope of the covariance kernel correlating modes. - n_integrate = 2_000 – Number of integration points for the harmonic transformation. - n_interpolate = 512 – Number of interpolation points for interpolating the kernel. - interpolation_dists_min_max – Interpolation range. Automatically chosen based on cutoff if left unspecified. - integration_dists_min_max – Integration range for the harmonic transform. Automatically chosen based on cutoff if left unspecified.

    The experimental matern kernel is deprecated and will be removed in a future version. We recommend using the new >MaternCovarianceModel< instead.

  • offset (tuple or callable or float) – Prior shift from zero in addition to the field intrinsic random shift.

  • window_size – Shape of the refinement matrices. Its value is automatically inferred from the padding in the grid or set to sane defaults if left unspecified.

  • compress (dict) – Arguments for compressing the refinement matrices. If this is set to an empty dictionary, no compression is applied.

  • prexi (str) – Prefix for the latent parameter names.

compress: bool#
covariance: Model | Callable#
fixed_kernel: bool#
grid: Grid#
kernel: Kernel#
offset: Model#