module: utils

This module is a collection of helpful functions.

annotlib.utils.check_indices(indices, max_index, parameter_name='indices')

This function checks whether the given indices are valid.

Parameters:

indices: array-like, shape (n_indices)

Array of indices to test.

max_index: int

The maximal allowed index.

parameter_name: str,

The name of the indices array, which is printed in case of error.

Returns:

indices: numpy.ndarray, shape (n_indices)

Is returned, if 0<= indices[i] <= max_index for all i and if indices contains no duplicates.

annotlib.utils.check_labelling_array(arr, shape, parameter_name='arr')

This function checks whether the given array has the given shape and all its values are in the interval [0, 1].

Parameters:

arr: array-like

Array whose shape is checked.

shape: array-like

The expected shape of the given array arr.

parameter_name: str,

The name of the indices array, which is printed in case of error.

Returns:

arr: array-like

Array whose shape is checked.

annotlib.utils.check_positive_integer(value, parameter_name='value')

This function checks whether the given value is a positive integer.

Parameters:

value: numeric,

Value to check.

parameter_name: str,

The name of the indices array, which is printed in case of error.

Returns:

value: int,

Checked and converted int.

annotlib.utils.check_range(arr, min_value, max_value, parameter_name='arr')

This function tests whether all elements of an array are in a given range.

Parameters:

arr: array-like,

Array.

min_value: float

Minimal number.

max_value: float

Maximal number.

parameter_name: str,

The name of the array arr, which is printed in case of error.

Returns:

arr: numpy.ndarray

Is returned, if min <= arr[i] <= max for all arr[i].

annotlib.utils.check_shape(arr, shape, parameter_name='arr')

This function checks whether the given array has the expected shape.

Parameters:

arr: array-like

Array whose shape is checked.

shape: array-like

The expected shape of the given array arr.

parameter_name: str,

The name of the indices array, which is printed in case of error.

Returns:

arr: array-like

Array whose shape is checked.

annotlib.utils.transform_confidences(C, n_classes)

Originally, non-adversarial annotators provide confidences in the interval [1/n_classes, 1]. In contrast, adversarial annotators have confidences in [0, 1/n_classes]. This function transforms the confidences into an interval [0, 1] for both annotator types. However, the meaning of the confidences is contrary. A class label provided by an adversarial annotator with confidence 1 is wrong in any case, whereas a class label provided by a non-adversarial annotator with confidence 1 is true in any case.

Parameters:

C: array-like, shape (n_samples, n_annotators)

confidences to be transformed.

n_classes: int

Number of classes.

Returns:

C_trans: array-like, shape (n_samples, n_annotators)

Transformed confidences.