Skip to content

Backward Method

Functions

backward(model, X, Y, loss)

Perform backward propagation through the entire model.

Parameters:

Name Type Description Default
model Sequential

The neural network model.

 required
X ndarray

Input data.

 required
Y ndarray

True labels.

 required
loss str

Loss function used.

 required
Example 
backward(model, X_train, Y_train, 'categorical_crossentropy')
Note

This function updates the gradients (dZ, dW, db) for all layers in the model.

Source code in microkeras/operations/backward/backward.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
27
28
29
30
31
32
33
34
def backward(model, X, Y, loss):
    """
    Perform backward propagation through the entire model.

    Args:
        model (Sequential): The neural network model.
        X (numpy.ndarray): Input data.
        Y (numpy.ndarray): True labels.
        loss (str): Loss function used.

    Example:
        ```python
        backward(model, X_train, Y_train, 'categorical_crossentropy')
        ```

    Note:
        This function updates the gradients (dZ, dW, db) for all layers in the model.
    """
    m = X.shape[1]  # number of training examples
    n_layers = len(model.layers)

    for i in reversed(range(n_layers)):
        if i == n_layers - 1:
            A_prev = model.layers[i-1].A if i > 0 else X
        else:
            A_prev = model.layers[i-1].A

        dZ, dW, db = backward_layer(model, i, X, Y, A_prev, loss, m)

        model.layers[i].dZ = dZ
        model.layers[i].dW = dW
        model.layers[i].db = db