Watch talks from the 2019 TensorFlow Dev Summit Watch now

tf.contrib.layers.l1_l2_regularizer

tf.contrib.layers.l1_l2_regularizer(
    scale_l1=1.0,
    scale_l2=1.0,
    scope=None
)

Defined in tensorflow/contrib/layers/python/layers/regularizers.py.

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

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.