使用tf.Variable()和tf.get_variable()时的结果不同 [英] Different outcomes when using tf.Variable() and tf.get_variable()

问题描述

我正在尝试通过此网站熟悉TensorFlow框架通过玩线性回归(LR).可以在此处中找到LR的源代码，名称03_linear_regression_sol.py.

I'm trying to get familiar with TensorFlow framework from this site by playing around with Linear Regression (LR). The source code for LR can be found here, with the name 03_linear_regression_sol.py.

通常，LR的定义模型为Y_predicted = X * w + b，其中

Generally, the defined model for LR is Y_predicted = X * w + b, where

• w和b是参数(tf.Variable)
• Y_predicted和X是训练数据(placeholder)

• w and b are parameters (tf.Variable)
• Y_predicted and X are training data (placeholder)

对于w和b，在示例代码中，它们的定义如下

For w and b, in the sample code, they are defined as following

w = tf.Variable(0.0, name='weights')
b = tf.Variable(0.0, name='bias')

我如下更改了这两行代码

And I changed these two lines of code a little bit as following

w = tf.get_variable('weights', [], dtype=tf.float32)
b = tf.get_variable('bias', [], dtype=tf.float32)

对于本实验，对于这两个版本，我得到了两个不同的total_loss/n_samples.更具体地说，在原始版本中，我随时可以得到确定性的结果，1539.0050282141283.但是，在修改后的版本中，我在不同的运行时间得到了不确定的结果，例如，total_loss/n_samples可能是1531.3039793868859，1526.3752814714044，...等.

For this experiment, I got two different total_loss/n_samples for those two versions. More specifically, in the original version, I got a deterministic result at anytime, 1539.0050282141283. But, in the modified version, I got undeterministic results at different running time, for example, total_loss/n_samples could be 1531.3039793868859, 1526.3752814714044, ... etc.

tf.Variable()和tf.get_variable()有什么区别?

推荐答案

tf.Variable 在创建时接受一个初始值(一个常数)，这解释了使用时的确定性结果.

tf.Variable accepts an initial value upon creation (a constant), this explains deterministic results when you use it.

tf.get_variable 略有不同:它具有initializer参数，默认情况下为None，其解释如下:

tf.get_variable is slightly different: it has an initializer argument, by default None, which is interpreted like this:

如果initializer是None(默认值)，则将使用在变量范围中传递的默认初始化程序.如果该名称也是None，则将使用glorot_uniform_initializer.

If initializer is None (the default), the default initializer passed in the variable scope will be used. If that one is None too, a glorot_uniform_initializer will be used.

由于您没有传递初始化程序，因此该变量具有统一的随机初始值.

Since you didn't pass an initializer, the variable got uniform random initial value.

这篇关于使用tf.Variable()和tf.get_variable()时的结果不同的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！