Home | History | Annotate | Download | only in keras 
      1 # Copyright 2015 The TensorFlow Authors. All Rights Reserved.
      2 #
      3 # Licensed under the Apache License, Version 2.0 (the "License");
      4 # you may not use this file except in compliance with the License.
      5 # You may obtain a copy of the License at
      6 #
      7 #     http://www.apache.org/licenses/LICENSE-2.0
      8 #
      9 # Unless required by applicable law or agreed to in writing, software
     10 # distributed under the License is distributed on an "AS IS" BASIS,
     11 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
     12 # See the License for the specific language governing permissions and
     13 # limitations under the License.
     14 # ==============================================================================
     15 """Built-in activation functions.
     16 """
     17 from __future__ import absolute_import
     18 from __future__ import division
     19 from __future__ import print_function
     20 
     21 import six
     22 
     23 from tensorflow.python.keras._impl.keras import backend as K
     24 from tensorflow.python.keras._impl.keras.utils.generic_utils import deserialize_keras_object
     25 from tensorflow.python.layers.base import Layer
     26 from tensorflow.python.platform import tf_logging as logging
     27 from tensorflow.python.util.tf_export import tf_export
     28 
     29 
     30 @tf_export('keras.activations.softmax')
     31 def softmax(x, axis=-1):
     32   """Softmax activation function.
     33 
     34   Arguments:
     35       x : Tensor.
     36       axis: Integer, axis along which the softmax normalization is applied.
     37 
     38   Returns:
     39       Tensor, output of softmax transformation.
     40 
     41   Raises:
     42       ValueError: In case `dim(x) == 1`.
     43   """
     44   ndim = K.ndim(x)
     45   if ndim == 2:
     46     return K.softmax(x)
     47   elif ndim > 2:
     48     e = K.exp(x - K.max(x, axis=axis, keepdims=True))
     49     s = K.sum(e, axis=axis, keepdims=True)
     50     return e / s
     51   else:
     52     raise ValueError('Cannot apply softmax to a tensor that is 1D')
     53 
     54 
     55 @tf_export('keras.activations.elu')
     56 def elu(x, alpha=1.0):
     57   return K.elu(x, alpha)
     58 
     59 
     60 @tf_export('keras.activations.selu')
     61 def selu(x):
     62   """Scaled Exponential Linear Unit. (Klambauer et al., 2017).
     63 
     64   Arguments:
     65       x: A tensor or variable to compute the activation function for.
     66 
     67   Returns:
     68       Tensor with the same shape and dtype as `x`.
     69 
     70   # Note
     71       - To be used together with the initialization "lecun_normal".
     72       - To be used together with the dropout variant "AlphaDropout".
     73 
     74   """
     75   alpha = 1.6732632423543772848170429916717
     76   scale = 1.0507009873554804934193349852946
     77   return scale * K.elu(x, alpha)
     78 
     79 
     80 @tf_export('keras.activations.softplus')
     81 def softplus(x):
     82   return K.softplus(x)
     83 
     84 
     85 @tf_export('keras.activations.softsign')
     86 def softsign(x):
     87   return K.softsign(x)
     88 
     89 
     90 @tf_export('keras.activations.relu')
     91 def relu(x, alpha=0., max_value=None):
     92   return K.relu(x, alpha=alpha, max_value=max_value)
     93 
     94 
     95 @tf_export('keras.activations.tanh')
     96 def tanh(x):
     97   return K.tanh(x)
     98 
     99 
    100 @tf_export('keras.activations.sigmoid')
    101 def sigmoid(x):
    102   return K.sigmoid(x)
    103 
    104 
    105 @tf_export('keras.activations.hard_sigmoid')
    106 def hard_sigmoid(x):
    107   return K.hard_sigmoid(x)
    108 
    109 
    110 @tf_export('keras.activations.linear')
    111 def linear(x):
    112   return x
    113 
    114 
    115 @tf_export('keras.activations.serialize')
    116 def serialize(activation):
    117   return activation.__name__
    118 
    119 
    120 @tf_export('keras.activations.deserialize')
    121 def deserialize(name, custom_objects=None):
    122   return deserialize_keras_object(
    123       name,
    124       module_objects=globals(),
    125       custom_objects=custom_objects,
    126       printable_module_name='activation function')
    127 
    128 
    129 @tf_export('keras.activations.get')
    130 def get(identifier):
    131   if identifier is None:
    132     return linear
    133   if isinstance(identifier, six.string_types):
    134     identifier = str(identifier)
    135     return deserialize(identifier)
    136   elif callable(identifier):
    137     if isinstance(identifier, Layer):
    138       logging.warning(
    139           'Do not pass a layer instance (such as {identifier}) as the '
    140           'activation argument of another layer. Instead, advanced '
    141           'activation layers should be used just like any other '
    142           'layer in a model.'.format(identifier=identifier.__class__.__name__))
    143     return identifier
    144   else:
    145     raise ValueError('Could not interpret '
    146                      'activation function identifier:', identifier)
    147