Coverage for /pythoncovmergedfiles/medio/medio/usr/local/lib/python3.8/site-packages/tensorflow/python/training/proximal_gradient_descent.py: 52%
27 statements
« prev ^ index » next coverage.py v7.4.0, created at 2024-01-03 07:57 +0000
« prev ^ index » next coverage.py v7.4.0, created at 2024-01-03 07:57 +0000
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# ==============================================================================
16"""ProximalGradientDescent for TensorFlow."""
17from tensorflow.python.framework import ops
18# pylint: disable=unused-import
19from tensorflow.python.ops import math_ops
20# pylint: enable=unused-import
21from tensorflow.python.training import optimizer
22from tensorflow.python.training import training_ops
23from tensorflow.python.util.tf_export import tf_export
26@tf_export(v1=["train.ProximalGradientDescentOptimizer"])
27class ProximalGradientDescentOptimizer(optimizer.Optimizer):
28 # pylint: disable=line-too-long
29 """Optimizer that implements the proximal gradient descent algorithm.
31 References:
32 Efficient Learning using Forward-Backward Splitting:
33 [Duchi et al., 2009](http://papers.nips.cc/paper/3793-efficient-learning-using-forward-backward-splitting)
34 ([pdf](http://papers.nips.cc/paper/3793-efficient-learning-using-forward-backward-splitting.pdf))
35 """
37 def __init__(self, learning_rate, l1_regularization_strength=0.0,
38 l2_regularization_strength=0.0, use_locking=False,
39 name="ProximalGradientDescent"):
40 """Construct a new proximal gradient descent optimizer.
42 Args:
43 learning_rate: A Tensor or a floating point value. The learning
44 rate to use.
45 l1_regularization_strength: A float value, must be greater than or
46 equal to zero.
47 l2_regularization_strength: A float value, must be greater than or
48 equal to zero.
49 use_locking: If True use locks for update operations.
50 name: Optional name prefix for the operations created when applying
51 gradients. Defaults to "GradientDescent".
52 """
53 super(ProximalGradientDescentOptimizer, self).__init__(use_locking, name)
54 self._learning_rate = learning_rate
55 self._l1_regularization_strength = l1_regularization_strength
56 self._l2_regularization_strength = l2_regularization_strength
57 self._l1_regularization_strength_tensor = None
58 self._l2_regularization_strength_tensor = None
60 def _apply_dense(self, grad, var):
61 return training_ops.apply_proximal_gradient_descent(
62 var,
63 self._learning_rate_tensor,
64 self._l1_regularization_strength_tensor,
65 self._l2_regularization_strength_tensor,
66 grad,
67 use_locking=self._use_locking).op
69 def _resource_apply_dense(self, grad, var):
70 return training_ops.resource_apply_proximal_gradient_descent(
71 var.handle,
72 self._learning_rate_tensor,
73 self._l1_regularization_strength_tensor,
74 self._l2_regularization_strength_tensor,
75 grad,
76 use_locking=self._use_locking)
78 def _apply_sparse(self, grad, var):
79 return training_ops.sparse_apply_proximal_gradient_descent(
80 var,
81 self._learning_rate_tensor,
82 self._l1_regularization_strength_tensor,
83 self._l2_regularization_strength_tensor,
84 grad.values,
85 grad.indices,
86 use_locking=self._use_locking).op
88 def _resource_apply_sparse(self, grad, var, indices):
89 return training_ops.resource_sparse_apply_proximal_gradient_descent(
90 var.handle,
91 math_ops.cast(self._learning_rate_tensor, grad.dtype),
92 math_ops.cast(self._l1_regularization_strength_tensor, grad.dtype),
93 math_ops.cast(self._l2_regularization_strength_tensor, grad.dtype),
94 grad,
95 indices,
96 use_locking=self._use_locking)
98 def _prepare(self):
99 self._learning_rate_tensor = ops.convert_to_tensor(self._learning_rate,
100 name="learning_rate")
101 self._l1_regularization_strength_tensor = ops.convert_to_tensor(
102 self._l1_regularization_strength, name="l1_regularization_strength")
103 self._l2_regularization_strength_tensor = ops.convert_to_tensor(
104 self._l2_regularization_strength, name="l2_regularization_strength")