kerasR (version 0.8.1)

Regularizers: Apply penalties on layer parameters

Description

Regularizers allow to apply penalties on layer parameters or layer activity during optimization. These penalties are incorporated in the loss function that the network optimizes.

Usage

l1(l = 0.01)
l2(l = 0.01)
l1_l2(l1 = 0.01, l2 = 0.01)

Arguments

l

multiplicitive factor to apply to the the penalty term

l1

multiplicitive factor to apply to the the l1 penalty term

l2

multiplicitive factor to apply to the the l2 penalty term

Author

Taylor B. Arnold,

Details

The penalties are applied on a per-layer basis. The exact API will depend on the layer, but the layers Dense, Conv1D, Conv2D and Conv3D have a unified API.

References

Chollet, Francois. 2015. Keras: Deep Learning library for Theano and TensorFlow.

Examples

if(keras_available()) {
  X_train <- matrix(rnorm(100 * 10), nrow = 100)
  Y_train <- to_categorical(matrix(sample(0:2, 100, TRUE), ncol = 1), 3)
  
  mod <- Sequential()
  mod$add(Dense(units = 50, input_shape = dim(X_train)[2]))
  mod$add(Activation("relu"))
  mod$add(Dense(units = 3, kernel_regularizer = l1(l = 0.05),
                bias_regularizer = l2(l = 0.05)))
  mod$add(Dense(units = 3, kernel_regularizer = l1_l2(l1 = 0.05, l2 = 0.1)))
  mod$add(Activation("softmax"))
  keras_compile(mod,  loss = 'categorical_crossentropy', optimizer = RMSprop())
  
  keras_fit(mod, X_train, Y_train, batch_size = 32, epochs = 5, verbose = 0)
}