pyinterp.fill.loess

Contents

pyinterp.fill.loess#

pyinterp.fill.loess(data, config=None, *, nx=None, ny=None, max_iterations=None, epsilon=None, value_type=None, first_guess=None, is_periodic=None, num_threads=None)[source]#

Fill missing values using LOESS (locally weighted regression).

Parameters:

  • data (_FloatArrayT) – 2D grid with missing values (NaN)

  • config (Loess | None) – Configuration object (if provided, keyword args are ignored)

  • nx (int | None) – Window size in X direction

  • ny (int | None) – Window size in Y direction

  • max_iterations (int | None) – Maximum number of iterations. If the value is 1, a single-pass LOESS is performed, the first guess option is ignored.

  • epsilon (float | None) – Convergence criterion

  • value_type (Literal['all', 'defined', 'undefined'] | None) – Which values to use in regression

  • first_guess (Literal['zero', 'zonal_average'] | None) – Initial guess method

  • is_periodic (bool | None) – Whether to assume periodic boundaries

  • num_threads (int | None) – Number of threads (0 = auto)

Returns:

Filled grid

Return type:

_FloatArrayT

Examples

>>> filled_grid = loess(
...     data, nx=5, ny=5, max_iterations=10, value_type="defined"
... )
Contents