Module `streamauc.utils`

Functions

def auc(vals_x: numpy.ndarray, vals_y: numpy.ndarray) ‑> Union[numpy.ndarray, float]

Compute the approximate area under the curve.

This is a weak wrapper around np.trapz, ensuring that the integral is always positive, i.e. making it ignore the sorting order of the input numpy array.

Parameters

vals_x : np.ndarray: Must be squeezable to shape (-1,) or (-1, n_classes).
vals_y : np.ndarray: Must be squeezable to shape (-1,) or (-1, n_classes).

Returns

Union[np.ndarray, float]: Approximate AUC value. Either single float or np.ndarray of shape (vals_x.shape[1],).

def check_confusion_matrix_entries(*args)

Validate that input confusion matrix arrays are 2D, have the same shape, and are non-negative.

Parameters:

*args : np.ndarray Variable number of input arrays representing confusion matrix entries.

Raises:

AssertionError If input arrays are not 2D, do not have the same shape, or contain negative values.

def copy_docstring_from(source)

Decorator to copy the docstring from one function to another.

Parameters

source : function: The function from which to copy the docstring.

Returns

function: The decorated function with the copied docstring.

def onehot_encode(int_masks: numpy.ndarray, num_classes: int) ‑> numpy.ndarray

Convert integer masks to one-hot encoded masks. Optimized methods that does not explode in memory for larger input arrays.

Parameters

int_masks : np.ndarray: An array of integer class labels. Each element in the array is an integer representing a class label.
num_classes : int: The number of distinct classes. This will determine the depth of the one-hot encoded dimension.

Returns

np.ndarray: A one-hot encoded array with shape (n, num_classes), where n is the number of elements in int_masks. Each row in the output corresponds to a one-hot encoded version of the respective element in int_masks.

Classes

class AggregationMethod (value, names=None, *, module=None, qualname=None, type=None, start=1)

Enumeration for specifying the method of aggregating metrics in multi-class classification.

Attributes:

MICRO : str Micro-averaging method, which aggregates contributions from all classes to compute the average metric. MACRO : str Macro-averaging method, which computes the metric independently for each class and then takes the average. ONE_VS_ALL : str One-vs-all method, which treats (subsequently) each class as the positive class and all others as negative, computing metrics for each class in this manner.

Expand source code

class AggregationMethod(Enum):
    """
    Enumeration for specifying the method of aggregating metrics in
    multi-class classification.

    Attributes:
    ----------
    MICRO : str
        Micro-averaging method, which aggregates contributions from all classes
        to compute the average metric.
    MACRO : str
        Macro-averaging method, which computes the metric independently for
        each class and then takes the average.
    ONE_VS_ALL : str
        One-vs-all method, which treats (subsequently) each class as the
        positive class and all others as negative, computing metrics for
        each class in this manner.
    """

    MICRO = "MICRO"
    MACRO = "MACRO"
    ONE_VS_ALL = "ONE_VS_ALL"

Ancestors

enum.Enum

Class variables

var MACRO
var MICRO
var ONE_VS_ALL