Skip to content

Categorical Crossentropy

Functions

categorical_crossentropy(Y, Y_hat)

Calculate the categorical cross-entropy loss.

Parameters:

Name Type Description Default
Y ndarray

True labels in one-hot encoded format.

 required
Y_hat ndarray

Predicted probabilities for each class.

 required

Returns:

Name Type Description
float

The categorical cross-entropy loss.
Example 
Y = np.array([[1, 0], [0, 1]])
Y_hat = np.array([[0.9, 0.1], [0.2, 0.8]])
loss = categorical_crossentropy(Y, Y_hat)
print(loss)
Note

A small epsilon (1e-8) is added to the log to avoid numerical instability when Y_hat contains zero probabilities.

Source code in microkeras/losses/categorical_crossentropy.py 
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
def categorical_crossentropy(Y, Y_hat):
    """
    Calculate the categorical cross-entropy loss.

    Args:
        Y (numpy.ndarray): True labels in one-hot encoded format.
        Y_hat (numpy.ndarray): Predicted probabilities for each class.

    Returns:
        float: The categorical cross-entropy loss.

    Example:
        ```python
        Y = np.array([[1, 0], [0, 1]])
        Y_hat = np.array([[0.9, 0.1], [0.2, 0.8]])
        loss = categorical_crossentropy(Y, Y_hat)
        print(loss)
        ```

    Note:
        A small epsilon (1e-8) is added to the log to avoid numerical instability
        when Y_hat contains zero probabilities.
    """
    return -np.sum(Y * np.log(Y_hat + 1e-8))