ranch.learning package

Submodules

ranch.learning.preprocessing.normalize(input: Cube, axis: Literal['spatial'], training_subset: Cube | Map | Profile | None = None, alternative_mode: bool = False) → Tuple[Cube, Map, Map][source]
ranch.learning.preprocessing.normalize(input: Cube, axis: Literal['spectral'], training_subset: Cube | Map | Profile | None = None, alternative_mode: bool = False) → Tuple[Cube, Profile, Profile]
ranch.learning.preprocessing.normalize(input: Cube, axis: Literal['all'] = 'all', training_subset: Cube | Map | Profile | None = None, alternative_mode: bool = False) → Tuple[Cube, float, float]
ranch.learning.preprocessing.normalize(input: Map, axis: Literal['all'] = 'all', training_subset: Map | None = None, alternative_mode: bool = False) → Tuple[Map, float, float]
ranch.learning.preprocessing.normalize(input: Profile, axis: Literal['all'] = 'all', training_subset: Profile | None = None, alternative_mode: bool = False) → Tuple[Profile, float, float]: If alternative_mode is False, return a modified version of input with every values between 0 and 1. If alternative_mode is True, return a modified version of input with every values between -1 and 1.

ranch.learning.preprocessing.scale(input: Cube, axis: Literal['spatial'], training_subset: Cube | Map | Profile | None = None, alternative_mode: bool = False) → Tuple[Cube, Map][source]
ranch.learning.preprocessing.scale(input: Cube, axis: Literal['spectral'], training_subset: Cube | Map | Profile | None = None, alternative_mode: bool = False) → Tuple[Cube, Profile]
ranch.learning.preprocessing.scale(input: Cube, axis: Literal['all'] = 'all', training_subset: Cube | Map | Profile | None = None, alternative_mode: bool = False) → Tuple[Cube, float]
ranch.learning.preprocessing.scale(input: Map, axis: Literal['all'] = 'all', training_subset: Map | None = None, alternative_mode: bool = False) → Tuple[Map, float]
ranch.learning.preprocessing.scale(input: Profile, axis: Literal['all'] = 'all', training_subset: Map | None = None, alternative_mode: bool = False) → Tuple[Profile, float]: If alternative_mode is False, return a modified version of input where every value is between -1 and 1 If alternative_mode is True, raise a NotImplementedError

ranch.learning.preprocessing.standardize(input: Cube, axis: Literal['spatial'], training_subset: Cube | Map | Profile | None = None, alternative_mode: bool = False) → Tuple[Cube, Map, Map][source]
ranch.learning.preprocessing.standardize(input: Cube, axis: Literal['spectral'], training_subset: Cube | Map | Profile | None = None, alternative_mode: bool = False) → Tuple[Cube, Profile, Profile]
ranch.learning.preprocessing.standardize(input: Cube, axis: Literal['all'] = 'all', training_subset: Cube | Map | Profile | None = None, alternative_mode: bool = False) → Tuple[Cube, float, float]
ranch.learning.preprocessing.standardize(input: Map, axis: Literal['all'] = 'all', training_subset: Map | None = None, alternative_mode: bool = False) → Tuple[Map, float, float]
ranch.learning.preprocessing.standardize(input: Profile, axis: Literal['all'] = 'all', training_subset: Profile | None = None, alternative_mode: bool = False) → Tuple[Profile, float, float]: If alternative_mode is False, return a modified version of input with mean 0 and standart deviation 1. If alternative_mode is True, return a modified version of input with mean 0 but same standart deviation.

ranch.learning.preprocessing.unnormalize(input: Cube, a: Map | Profile | float, b: Map | Profile | float) → Cube[source]
ranch.learning.preprocessing.unnormalize(input: Map, a: float, b: float) → Map
ranch.learning.preprocessing.unnormalize(input: Profile, a: float, b: float) → Profile: TODO

ranch.learning.preprocessing.unscale(input: Cube, s: Map | Profile | float) → Cube[source]
ranch.learning.preprocessing.unscale(input: Map, s: float) → Map
ranch.learning.preprocessing.unscale(input: Profile, s: float) → Profile: TODO

ranch.learning.preprocessing.unstandardize(input: Cube, mu: Map | Profile | float, sigma: Map | Profile | float, alternative_mode: bool = False) → Cube[source]
ranch.learning.preprocessing.unstandardize(input: Map, mu: float, sigma: float, alternative_mode: bool = False) → Map
ranch.learning.preprocessing.unstandardize(input: Profile, mu: float, sigma: float, alternative_mode: bool = False) → Profile: TODO

ranch.learning.splitting.extract_indices(input: Map | Profile) → tuple[ndarray][source]: Returns the indices of the training, validation and full sets.

ranch.learning.splitting.nan_mask(input: Struct, item: str) → Struct[source]: Returns a binary mask equal to 1 if an item contains at least a NaN, else 0.

ranch.learning.splitting.random_spatial_splitting(inputs: Struct | Sequence[Struct], item: str)[source]: Returns a ternary structure equal to 1 if an item is in the training set, 0 if it is in the validation set and -1 if the item contains at least one NaN

ranch.learning.splitting.random_splitting(inputs: Struct | Sequence[Struct], item: str, frac_train: float, seed: int | None = None, reject_nans: bool = True) → Struct[source]: Returns a ternary structure equal to 1 if an item is in the training set, 0 if it is in the validation set and -1 if the item contains at least one NaN. The splitting structure is valid for every inputect in inputs.

ranch.learning.splitting.spatial_splitting(inputs: Struct | Sequence[Struct], item: str)[source]: Returns a ternary structure equal to 1 if an item is in the training set, 0 if it is in the validation set and -1 if the item contains at least one NaN

ranch.learning.splitting.stack_indices(extracted: Sequence[tuple[ndarray]]) → tuple[ndarray][source]: Returns the stacked indices of the training, validation and full sets.