Skip to content
体验新版
项目
组织
正在加载...
登录
切换导航
打开侧边栏
wzpei9
Paddle
提交
d2b791a0
P
Paddle
项目概览
wzpei9
/
Paddle
与 Fork 源项目一致
Fork自
PaddlePaddle / Paddle
通知
1
Star
1
Fork
0
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
1
列表
看板
标记
里程碑
合并请求
0
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
P
Paddle
项目概览
项目概览
详情
发布
仓库
仓库
文件
提交
分支
标签
贡献者
分支图
比较
Issue
1
Issue
1
列表
看板
标记
里程碑
合并请求
0
合并请求
0
Pages
分析
分析
仓库分析
DevOps
Wiki
0
Wiki
成员
成员
收起侧边栏
关闭侧边栏
动态
分支图
创建新Issue
提交
Issue看板
体验新版 GitCode,发现更多精彩内容 >>
提交
d2b791a0
编写于
6月 17, 2018
作者:
Q
qiaolongfei
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
add SGD and momentum optimizer doc
上级
16a0f746
变更
1
隐藏空白更改
内联
并排
Showing
1 changed file
with
57 addition
and
9 deletion
+57
-9
python/paddle/fluid/optimizer.py
python/paddle/fluid/optimizer.py
+57
-9
未找到文件。
python/paddle/fluid/optimizer.py
浏览文件 @
d2b791a0
...
...
@@ -28,8 +28,8 @@ from contextlib import contextmanager
__all__
=
[
'SGD'
,
'Momentum'
,
'Adagrad'
,
'Adam'
,
'Adamax'
,
'DecayedAdagrad'
,
'SGDOptimizer'
,
'MomentumOptimizer'
,
'AdagradOptimizer'
,
'AdamOptimizer'
,
'AdamaxOptimizer'
,
'DecayedAdagradOptimizer'
,
'
RMSProp
Optimizer'
,
'Adadelta'
,
'ModelAverage'
,
'Optimizer'
'AdamaxOptimizer'
,
'DecayedAdagradOptimizer'
,
'
Adadelta
Optimizer'
,
'
RMSPropOptimizer'
,
'
Adadelta'
,
'ModelAverage'
,
'Optimizer'
]
...
...
@@ -192,15 +192,15 @@ class Optimizer(object):
"""Add optimization operators to update gradients to variables.
Args:
loss: the target that this optimization is for.
parameters_and_grads: a list of (variable, gradient) pair to update.
loss(Variable): the target that this optimization is for.
parameters_and_grads(list(tuple(Variable, Variable))):
a list of (variable, gradient) pair to update.
Returns:
return_op_list: a list of operators that will complete one step of
optimization. This will include parameter update ops, global step
update ops and any other custom ops required by subclasses to manage
their internal state.
:param startup_program:
"""
# This is a default implementation of create_optimization_pass that
# can be shared by most optimizers. This implementation assumes that
...
...
@@ -268,7 +268,22 @@ class Optimizer(object):
class
SGDOptimizer
(
Optimizer
):
""" Simple SGD optimizer without any state.
"""
Optimizer of the stochastic gradient descent algorithm.
.. math::
param\_out = param - learning\_rate * grad
Args:
learning_rate (float|Variable): the learning rate used to update parameters.
\
Can be a float value or a Variable with one float value as data element.
Examples:
.. code-block:: python
sgd_optimizer = SGDOptimizer(learning_rate=0.2)
sgd_optimizer.minimize(cost)
"""
def
__init__
(
self
,
learning_rate
,
**
kwargs
):
...
...
@@ -294,7 +309,37 @@ class SGDOptimizer(Optimizer):
class
MomentumOptimizer
(
Optimizer
):
"""Simple Momentum optimizer with velocity state
"""
Simple Momentum optimizer with velocity state
This optimizer has a flag for Nestrov Momentum.
The update equations are as follows:
.. math::
& velocity = mu * velocity + gradient
& if (use\_nesterov):
& param = param - gradient * learning\_rate + mu * velocity * learning\_rate
& else:
& param = param - learning\_rate * velocity
Args:
learning_rate (float|Variable): the learning rate used to update parameters.
\
Can be a float value or a Variable with one float value as data element.
momentum (float): momentum factor
use_nesterov (bool): enables Nesterov momentum
Examples:
.. code-block:: python
optimizer = MomentumOptimizer(learning_rate=0.2, momentum=0.1)
optimizer.minimize(cost)
"""
_velocity_acc_str
=
"velocity"
...
...
@@ -614,6 +659,7 @@ class DecayedAdagradOptimizer(Optimizer):
class
AdadeltaOptimizer
(
Optimizer
):
"""
**Adadelta Optimizer**
Simple Adadelta optimizer with average squared grad state and
average squared update state.
The details of adadelta please refer to this
...
...
@@ -703,7 +749,7 @@ class RMSPropOptimizer(Optimizer):
.. math::
r(w, t) & =
\\
rho r(w, t-1) + (1 -
\\
rho)(
\\
nabla Q_{i}(w))^2
\\
\\
r(w, t) & =
\\
rho r(w, t-1) + (1 -
\\
rho)(
\\
nabla Q_{i}(w))^2
\\
w & = w -
\\
frac{
\\
eta} {
\\
sqrt{r(w,t) +
\\
epsilon}}
\\
nabla Q_{i}(w)
...
...
@@ -844,7 +890,9 @@ class ModelAverage(Optimizer):
max_average_window: The maximum size of average window.
Examples:
...
.. code-block:: python
optimizer = fluid.optimizer.Momentum()
_, params_grads = optimizer.minimize(cost)
model_average = fluid.optimizer.ModelAverage(params_grads, 0.15,
...
...
编辑
预览
Markdown
is supported
0%
请重试
或
添加新附件
.
添加附件
取消
You are about to add
0
people
to the discussion. Proceed with caution.
先完成此消息的编辑!
取消
想要评论请
注册
或
登录