Deep Learning with Keras

Regularization

Regularization is a way to prevent overfitting. We have already seen examples of usage in Chapter 1, Neural Networks Foundations. Multiple layers have parameters for regularization.  The following is the list of regularization parameters commonly used for dense, and convolutional modules:

• kernel_regularizer: Regularizer function applied to the weight matrix
• bias_regularizer: Regularizer function applied to the bias vector
• activity_regularizer: Regularizer function applied to the output of the layer (its activation)

In addition is possible to use Dropout for regularization and that is frequently a very effective choice

keras.layers.core.Dropout(rate, noise_shape=None, seed=None)

Where:

• rate: It is a float between 0 and ...