TensorFlow 2.0 Beta is available Learn more

tf.contrib.layers.l1_l2_regularizer

Returns a function that can be used to apply L1 L2 regularizations.

tf.contrib.layers.l1_l2_regularizer(
    scale_l1=1.0,
    scale_l2=1.0,
    scope=None
)
View source on GitHub

Args:

  • scale_l1: A scalar multiplier Tensor for L1 regularization.
  • scale_l2: A scalar multiplier Tensor for L2 regularization.
  • scope: An optional scope name.

Returns:

A function with signature l1_l2(weights) that applies a weighted sum of L1 L2 regularization.

Raises:

  • ValueError: If scale is negative or if scale is not a float.