tf.Variable

TF.Variable

1.前言

在学习莫烦Tensorflow教学视频中有些东西不清楚，这篇是关于tf.Variable的拓展。

2.变量的创建

视频中用的tf.Variable(tf.zeros([1]))可以简写为tf.zeros([1])。
2.1生成Tensor

tf.zeros(shape, dtype=tf.float32, name=None)
tf.zeros_like(tensor, dtype=None, name=None)
tf.constant(value, dtype=None, shape=None, name='Const')
tf.fill(dims, value, name=None)
tf.ones_like(tensor, dtype=None, name=None)
tf.ones(shape, dtype=tf.float32, name=None)

2.2生成序列
类似于Matlab中的用法

tf.range(start, limit, delta=1, name='range')
tf.linspace(start, stop, num, name=None)

2.3生成随机矩阵

tf.random_normal(shape, mean=0.0, stddev=1.0,
dtype=tf.float32, seed=None, name=None)
tf.truncated_normal(shape, mean=0.0, stddev=1.0, dtype=tf.float32, seed=None, name=None)
tf.random_uniform(shape, minval=0.0, maxval=1.0, dtype=tf.float32, seed=None, name=None)
tf.random_shuffle(value, seed=None, name=None)

2.4.1例子

import tensorflow as tf

# 生成0和1矩阵
v1 = tf.Variable(tf.zeros([3, 3, 3]), name="v1")
v2 = tf.Variable(tf.ones([2, 3]), name="v2")

# 填充单值矩阵
v3 = tf.Variable(tf.fill([2, 3], 9))

# 常量矩阵
v4_1 = tf.constant([1, 2, 3, 4, 5, 6, 7])
v4_2 = tf.constant(-1.0, shape=[2, 3])

# 生成等差数列
v6_1 = tf.linspace(1.0, 10.0, 20, name="linspace")  # float32 or float64
v7_1 = tf.range(10, 20, 3)  # just int32

# 生成各种随机数据矩阵
# seed 为标准差
v8_1 = tf.random_uniform([2, 4], minval=0.0, maxval=2.0, dtype=tf.float32, seed=1234, name="v8_1")
v8_2 = tf.Variable(tf.random_normal([2, 3], mean=0.0, stddev=1.0, dtype=tf.float32, seed=1234, name="v8_2"))
v8_3 = tf.Variable(tf.truncated_normal([2, 3], mean=0.0, stddev=1.0, dtype=tf.float32, seed=1234, name="v8_3"))
v8_4 = tf.Variable(tf.random_uniform([2, 3], minval=0.0, maxval=1.0, dtype=tf.float32, seed=1234, name="v8_4"))
v8_5 = tf.random_shuffle([[1, 2, 3], [4, 5, 6], [6, 6, 6]], seed=134, name="v8_5")

# 初始化
init_op = tf.global_variables_initializer()


# 运行
with tf.Session() as sess:
    sess.run(init_op)
    print('v1')
    print(v1)
    print(tf.Variable.get_shape(v1))
    print(sess.run(v1))

    print('v2')
    print(sess.run(v2))

    print('v3')
    print(sess.run(v3))

    print('v4_1')
    print(sess.run(v4_1))
    print('v4_2')
    print(sess.run(v4_2))

    print('v6_1')
    print(sess.run(v6_1))
    print('v7_1')
    print(sess.run(v7_1))

    print('v8_1')
    print(sess.run(v8_1))
    print('v8_2')
    print(sess.run(v8_2))
    print('v8_3')
    print(sess.run(v8_3))
    print('v8_4')
    print(sess.run(v8_4))
    print('v8_5')
    print(sess.run(v8_5))

2.4.2例子

import tensorflow as tf

state = tf.Variable(0, name='counter')
#print(state.name)
one = tf.constant(1)

new_value = tf.add(state, one)
update = tf.assign(state, new_value) # 赋值操作 将new_value赋值给state ， 注意updata为操作。

init = tf.initialize_all_variables()  # must have if define variable

with tf.Session() as sess:
    sess.run(init)
    for _ in range(3):
        sess.run(update)
        print(sess.run(state))

3.变量的初始化

变量的初始化必须在模型的其它操作运行之前先明确地完成。
3.1最简单的方法
添加一个给所有变量初始化的操作，并在使用模型之前首先运行那个操作。

# Create two variables.
weights = tf.Variable(tf.random_normal([784, 200], stddev=0.35),
                      name="weights")
biases = tf.Variable(tf.zeros([200]), name="biases")
...
# Add an op to initialize the variables.
init_op = tf.global_variables_initializer()
# tf.initialize_all_variables()这是旧版本的用法

# Later, when launching the model
with tf.Session() as sess:
  # Run the init operation.
  sess.run(init_op)
  ...
  # Use the model
  ...

4.引用

liumw1203的简书
修雨轩陈的Blog