File: optimizers.md

package info (click to toggle)
keras 2.3.1%2Bdfsg-3
  • links: PTS, VCS
  • area: main
  • in suites: bullseye
  • size: 9,288 kB
  • sloc: python: 48,266; javascript: 1,794; xml: 297; makefile: 36; sh: 30
file content (238 lines) | stat: -rw-r--r-- 6,727 bytes parent folder | download
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238

## Usage of optimizers

An optimizer is one of the two arguments required for compiling a Keras model:

```python
from keras import optimizers

model = Sequential()
model.add(Dense(64, kernel_initializer='uniform', input_shape=(10,)))
model.add(Activation('softmax'))

sgd = optimizers.SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)
model.compile(loss='mean_squared_error', optimizer=sgd)
```

You can either instantiate an optimizer before passing it to `model.compile()` , as in the above example, or you can call it by its name. In the latter case, the default parameters for the optimizer will be used.

```python
# pass optimizer by name: default parameters will be used
model.compile(loss='mean_squared_error', optimizer='sgd')
```

---

## Parameters common to all Keras optimizers

The parameters `clipnorm` and `clipvalue` can be used with all optimizers to control gradient clipping:

```python
from keras import optimizers

# All parameter gradients will be clipped to
# a maximum norm of 1.
sgd = optimizers.SGD(lr=0.01, clipnorm=1.)
```

```python
from keras import optimizers

# All parameter gradients will be clipped to
# a maximum value of 0.5 and
# a minimum value of -0.5.
sgd = optimizers.SGD(lr=0.01, clipvalue=0.5)
```

---

<span style="float:right;">[[source]](https://github.com/keras-team/keras/blob/master/keras/optimizers.py#L164)</span>
### SGD

```python
keras.optimizers.SGD(learning_rate=0.01, momentum=0.0, nesterov=False)
```

Stochastic gradient descent optimizer.

Includes support for momentum,
learning rate decay, and Nesterov momentum.

__Arguments__

- __learning_rate__: float >= 0. Learning rate.
- __momentum__: float >= 0. Parameter that accelerates SGD
    in the relevant direction and dampens oscillations.
- __nesterov__: boolean. Whether to apply Nesterov momentum.
    
----

<span style="float:right;">[[source]](https://github.com/keras-team/keras/blob/master/keras/optimizers.py#L229)</span>
### RMSprop

```python
keras.optimizers.RMSprop(learning_rate=0.001, rho=0.9)
```

RMSProp optimizer.

It is recommended to leave the parameters of this optimizer
at their default values
(except the learning rate, which can be freely tuned).

__Arguments__

- __learning_rate__: float >= 0. Learning rate.
- __rho__: float >= 0.

__References__

- [rmsprop: Divide the gradient by a running average of its recent magnitude
   ](http://www.cs.toronto.edu/~tijmen/csc321/slides/lecture_slides_lec6.pdf)
    
----

<span style="float:right;">[[source]](https://github.com/keras-team/keras/blob/master/keras/optimizers.py#L303)</span>
### Adagrad

```python
keras.optimizers.Adagrad(learning_rate=0.01)
```

Adagrad optimizer.

Adagrad is an optimizer with parameter-specific learning rates,
which are adapted relative to how frequently a parameter gets
updated during training. The more updates a parameter receives,
the smaller the learning rate.

It is recommended to leave the parameters of this optimizer
at their default values.

__Arguments__

- __learning_rate__: float >= 0. Initial learning rate.

__References__

- [Adaptive Subgradient Methods for Online Learning and Stochastic
   Optimization](http://www.jmlr.org/papers/volume12/duchi11a/duchi11a.pdf)
    
----

<span style="float:right;">[[source]](https://github.com/keras-team/keras/blob/master/keras/optimizers.py#L376)</span>
### Adadelta

```python
keras.optimizers.Adadelta(learning_rate=1.0, rho=0.95)
```

Adadelta optimizer.

Adadelta is a more robust extension of Adagrad
that adapts learning rates based on a moving window of gradient updates,
instead of accumulating all past gradients. This way, Adadelta continues
learning even when many updates have been done. Compared to Adagrad, in the
original version of Adadelta you don't have to set an initial learning
rate. In this version, initial learning rate and decay factor can
be set, as in most other Keras optimizers.

It is recommended to leave the parameters of this optimizer
at their default values.

__Arguments__

- __learning_rate__: float >= 0. Initial learning rate, defaults to 1.
    It is recommended to leave it at the default value.
- __rho__: float >= 0. Adadelta decay factor, corresponding to fraction of
    gradient to keep at each time step.

__References__

- [Adadelta - an adaptive learning rate method](
   https://arxiv.org/abs/1212.5701)
    
----

<span style="float:right;">[[source]](https://github.com/keras-team/keras/blob/master/keras/optimizers.py#L467)</span>
### Adam

```python
keras.optimizers.Adam(learning_rate=0.001, beta_1=0.9, beta_2=0.999, amsgrad=False)
```

Adam optimizer.

Default parameters follow those provided in the original paper.

__Arguments__

- __learning_rate__: float >= 0. Learning rate.
- __beta_1__: float, 0 < beta < 1. Generally close to 1.
- __beta_2__: float, 0 < beta < 1. Generally close to 1.
- __amsgrad__: boolean. Whether to apply the AMSGrad variant of this
    algorithm from the paper "On the Convergence of Adam and
    Beyond".

__References__

- [Adam - A Method for Stochastic Optimization](
   https://arxiv.org/abs/1412.6980v8)
- [On the Convergence of Adam and Beyond](
   https://openreview.net/forum?id=ryQu7f-RZ)
    
----

<span style="float:right;">[[source]](https://github.com/keras-team/keras/blob/master/keras/optimizers.py#L567)</span>
### Adamax

```python
keras.optimizers.Adamax(learning_rate=0.002, beta_1=0.9, beta_2=0.999)
```

Adamax optimizer from Adam paper's Section 7.

It is a variant of Adam based on the infinity norm.
Default parameters follow those provided in the paper.

__Arguments__

- __learning_rate__: float >= 0. Learning rate.
- __beta_1__: float, 0 < beta < 1. Generally close to 1.
- __beta_2__: float, 0 < beta < 1. Generally close to 1.

__References__

- [Adam - A Method for Stochastic Optimization](
   https://arxiv.org/abs/1412.6980v8)
    
----

<span style="float:right;">[[source]](https://github.com/keras-team/keras/blob/master/keras/optimizers.py#L645)</span>
### Nadam

```python
keras.optimizers.Nadam(learning_rate=0.002, beta_1=0.9, beta_2=0.999)
```

Nesterov Adam optimizer.

Much like Adam is essentially RMSprop with momentum,
Nadam is RMSprop with Nesterov momentum.

Default parameters follow those provided in the paper.
It is recommended to leave the parameters of this optimizer
at their default values.

__Arguments__

- __learning_rate__: float >= 0. Learning rate.
- __beta_1__: float, 0 < beta < 1. Generally close to 1.
- __beta_2__: float, 0 < beta < 1. Generally close to 1.

__References__

- [Nadam report](http://cs229.stanford.edu/proj2015/054_report.pdf)
- [On the importance of initialization and momentum in deep learning](
   http://www.cs.toronto.edu/~fritz/absps/momentum.pdf)