tensorflow-Constants, Sequences, Random Values

tf.random_uniform( shape, minval=0, maxval=None, dtype=tf.float32, seed=None, name=None )

均匀分布

#!/usr/bin/env python
# coding=utf-8
import numpy as np
import tensorflow as tf

var = tf.Variable(tf.random_uniform([2,3]), name="var" )
init=tf.initialize_all_variables()

sess=tf.Session()
sess.run(init)
print sess.run(var)
sess.close()

tf.truncated_normal(shape, mean=0.0, stddev=1.0, dtype=tf.float32, seed=None, name=None)

输出服从一个阶段正态分布的随机值

产生服务指定均值和标准差的正态分布的值，除了距离mean超过2倍标准差的值。

#!/user/bin/env python
# coding=utf-8

import tensorflow as tf

vec=tf.Variable( tf.truncated_normal([2,3],mean=0.0,stddev=1.0,dtype=tf.float32,seed=1234,name="vec"))
init=tf.initialize_all_variables()
sess=tf.Session()
sess.run(init)
print sess.run(vec)
sess.close()

tf.ones(shape, dtype=tf.float32, name=None)

返回一个tensor，所有的元素都为1

#!/usr/bin/env python
# coding=utf-8

import tensorflow as tf

one=tf.ones([2,3],tf.int32)

sess=tf.Session()
print sess.run(one)
sess.close()

tf.ones_like(tensor, dtype=None, name=None, optimize=True)

创建一个所有元素都为1的tensor

#!/usr/bin/env python
# coding=utf-8

import tensorflow as tf

one=tf.ones([2,3],tf.int32)
one_like=tf.ones_like(one,dtype=tf.float32)*0.5
sess=tf.Session()
print sess.run(one_like)
sess.close()

tf.greater(x,y,name=None)

返回元素级的真值

#!/usr/bin/env python
# coding=utf-8

import tensorflow as tf

gt=tf.greater([0.8,0.2],[0.5,0.5])

sess=tf.Session()
print sess.run(gt)
sess.close()

tf.cast(x,dtype, name=None)

将一个tensor转换为一个新的类型

#!/usr/bin/env python
# coding=utf-8

import tensorflow as tf

gt=tf.greater([0.8,0.2],[0.5,0.5])
t_gt=tf.cast(gt,tf.int32)
sess=tf.Session()
print sess.run([gt,t_gt])
sess.close()

tf.equal(x,y, name=None)

返回元素级的真值（x==y）

#!/usr/bin/env python
# coding=utf-8

import tensorflow as tf

gt=tf.greater([0.8,0.2],[0.5,0.5])
t_gt=tf.cast(gt,tf.int32)
equ=tf.equal([1,1],t_gt)
sess=tf.Session()
ret=sess.run(equ)
print ret
sess.close()

tf.argmax(input, dimension, name=None)

根据tensor的维数返回最大值所在的索引。

dimension: 0<=dimension<rank(input), 对于一个向量，dimension=0

#!/usr/bin/env python
# coding=utf-8

import tensorflow as tf
import numpy as np

y=np.array([0,0,0,0,0,1,0,0,0])
y_idx=tf.arg_max(y,0)
x=np.array([[0,0,1,0,0,0,0,0,0],[0,1,0,0,0,0,0,0,0],[0,0,0,0,0,0,0,0,1]])
x_idx=tf.arg_max(x,1)


sess=tf.Session()
print sess.run(y_idx)
print sess.run(x_idx)
sess.close()