import tensorflow as tf

w1=tf.Variable(tf.random_uniform([2,2],-1,1))

with tf.Session() as sess:
    init = tf.global_variables_initializer()
    sess.run(init)
    print(sess.run(w1))

輸出得

[[ 0.94702697 -0.19546461]
 [-0.69034481 -0.10538507]]

n_features=7

w2=tf.Variable(tf.random_normal((n_features,1),mean=0.0,stddev=1.0),name='weights')

with tf.Session() as sess:
    init = tf.global_variables_initializer()
    sess.run(init)
    print(sess.run(w2))
 

輸出：

[[-0.6331901 ]
 [ 2.41542029]
 [-0.83030796]
 [-1.55228913]
 [-0.98885322]
 [-0.63248575]
 [ 0.26187277]]

def feature_normalize(data):
    mu = np.mean(data,axis=0)
    std = np.std(data,axis=0)
    return (data - mu)/std

其中，np.std()的運算為

std = sqrt(mean(abs(x - x.mean())**2))

如果沒有指定axis，則是全部元素計算。

注意這裡涉及的向量運算是基於每個元素的運算。

std是標準差不是方差。

import numpy as np
ind = np.random.permutation(20)
print(ind)

輸出

[11 15  4  1  6 12  8 17 18 14  9  5  7  0 13 10 19  3 16  2]

import numpy as np
import matplotlib.pyplot as plt
s = np.random.poisson(5)
print(s)

s2 = np.random.poisson(5, 10000)
count, bins, ignored = plt.hist(s2, 14, density=True)
plt.show()
 

輸出：4