采用TensorFlow的截断式正太分布来生成数据

所谓的截断式正态分布就是在标准正态分（高斯分布）布的基础上加以限制，以使得生成的数据在一定范围上。如：标准正态分布生生成的数据在负无穷到正无穷，但是截断式正态分布生成的数据在（均值-2倍的标准差，均值+2倍的标准差）这个范围内。

import tensorflow as tf
c = tf.truncated_normal(shape=[12,10], mean=100, stddev=1)
with tf.Session() as sess:
	print (sess.run(c))

输出结果：

[[  99.91983795   98.71269226   99.35746002  101.02975464  100.63500977]
 [ 101.92388153   99.10471344   99.77385712  100.06600952   99.75480652]
 [  99.94233704  100.22845459   99.16555023  101.3265686    99.82016754]
 [ 100.27728271   99.2677002   100.80665588   99.7434845   100.93345642]
 [  99.00519562  100.91706848  100.16968536   99.20123291   99.35009766]
 [  98.85636902  100.11727142  100.32691193  100.59256744   99.72653961]
 [  99.51020813   98.99095917   99.38573456   99.34537506  100.68587494]
 [  99.56321716   99.81674194  100.38961029  100.29162598   98.17160034]
 [ 101.67827606  101.01641083  100.11600494  100.59056091   99.62102509]
 [  98.12446594   99.39179993   99.95746613   99.32868195  100.14764404]]

其中mean为均值，stddev是方差