1 #学习TensorFlow笔记 2 import tensorflow as tf 3 4 #定义变量 5 #Variable 定义张量及shape 6 w1= tf.Variable(tf.random_normal([2, 3], stddev=1, seed=1)) 7 w2= tf.Variable(tf.random_normal([3, 1], stddev=1, seed=1)) 8 with tf.Session() as sess: 9 print(sess.run(w1.initializer)) 10 print(sess.run(w2.initializer)) 11 #None 12 #None 13 14 15 16 #打印张量，查看数据shape等信息 17 print(w1) 18 print(w2) 19 #<tf.Variable 'Variable:0' shape=(2, 3) dtype=float32_ref> 20 #<tf.Variable 'Variable_1:0' shape=(3, 1) dtype=float32_ref> 21 22 #tf.constan是一个计算，结果为一个张量，保存在变量x中 23 x = tf.constant([[0.7, 0.9]]) 24 print(x) 25 #Tensor("Const:0", shape=(1, 2), dtype=float32) 26 with tf.Session() as sess: 27 print(sess.run(x)) 28 #[[ 0.69999999 0.89999998]] 29 30 31 #定义前向传播的神经网络 32 #matmul做矩阵乘法 33 a = tf.matmul(x, w1) # x shape=(1, 2) w1 shape=(2, 3) 34 35 print(a) 36 #Tensor("MatMul:0", shape=(1, 3), dtype=float32) 37 38 y = tf.matmul(a, w2) #a shape=(1, 3) w2 shape=(3, 1) 39 print(y) 40 #Tensor("MatMul_1:0", shape=(1, 1), dtype=float32) 41 42 43 #调用会话输出结果 44 with tf.Session() as sess: 45 sess.run(w1.initializer) 46 sess.run(w2.initializer) 47 print(sess.run(a)) 48 #[[-2.76635647 1.12854266 0.57783246]] 49 print(sess.run(y)) 50 #[[ 3.95757794]] 51 52 #placeholder 53 x=tf.placeholder(tf.float32,shape=(1,2),name="input") 54 a=tf.matmul(x,w1) 55 y=tf.matmul(a,w2) 56 sess=tf.Session() 57 init_op=tf.global_variables_initializer() 58 sess.run(init_op) 59 60 print(sess.run(y,feed_dict={x:[[0.8,0.9]]})) 61 #[[ 4.2442317]] 62 x = tf.placeholder(tf.float32, shape=(3, 2), name="input") 63 a = tf.matmul(x, w1) 64 y = tf.matmul(a, w2) 65 66 sess = tf.Session() 67 #使用tf.global_variables_initializer()来初始化所有的变量 68 init_op = tf.global_variables_initializer() 69 sess.run(init_op) 70 71 print(sess.run(y, feed_dict={x: [[0.7,0.9],[0.1,0.4],[0.5,0.8]]})) 72 73 ''' 74 [[ 3.95757794] 75 [ 1.15376544] 76 [ 3.16749239]] 77 ''' 78
#整体神经网络的实现 81 import tensorflow as tf 82 from numpy.random import RandomState 83 #定义神经网络的参数，输入和输出节点 84 batch_size=8 85 #均值为0 方差为1 随机分布满足正态分布 shape为2*3 86 w1=tf.Variable(tf.random_normal([2,3],stddev=1,seed=1)) 87 w2=tf.Variable(tf.random_normal([3, 1], stddev=1, seed=1)) 88 #shape 根据数据自动计算 batchsize个 89 x=tf.placeholder(tf.float32,shape=(None,2)) 90 y_=tf.placeholder(tf.float32,shape=(None,1)) 91 92 #定义前向传播过程，损失函数及反向传播算法 93 94 a=tf.matmul(x,w1) 95 y=tf.matmul(a,w2) 96 #损失函数 使用交叉熵 97 #优化方法使用AdamOptimizer 98 cross_entropy=-tf.reduce_mean(y_*tf.log(tf.clip_by_value(y,1e-10,1.0))) 99 train_step=tf.train.AdamOptimizer(learning_rate=0.001).minimize(cross_entropy) 100 101 rdm=RandomState(1) 102 #随机生成128个数据 shape 128*2 103 X=rdm.rand(128,2) 104 105 #Y的值是模拟的 ，实际假设x2+x1如果大于1则标签Y为1 否则标签Y为0 106 Y=[[int(x1+x2<1)] for (x1,x2) in X] 107 108 #创建一个会话 ，运算计算图 109 #全局初始化变量 110 STEPS = 5000 111 with tf.Session() as sess: 112 init_op=tf.global_variables_initializer() 113 sess.run(init_op) 114 # 输出目前（未经训练）的参数取值。 115 print("w1:", sess.run(w1)) 116 print("w2:", sess.run(w2)) 117 print(" ") 118 for i in range(STEPS): 119 start = (i * batch_size) % 128 120 end = (i * batch_size) % 128 + batch_size 121 sess.run(train_step, feed_dict={x: X[start:end], y_: Y[start:end]}) 122 if i % 1000 == 0: 123 total_cross_entropy = sess.run(cross_entropy, feed_dict={x: X, y_: Y}) 124 print("After %d training step(s), cross entropy on all data is %g" % (i, total_cross_entropy)) 125 # 输出训练后的参数取值。 126 print(" ") 127 print("w1:", sess.run(w1)) 128 print("w2:", sess.run(w2)) 129 130 ''' 131 132 w1: [[-0.81131822 1.48459876 0.06532937] 133 [-2.4427042 0.0992484 0.59122431]] 134 w2: [[-0.81131822] 135 [ 1.48459876] 136 [ 0.06532937]] 137 138 139 After 0 training step(s), cross entropy on all data is 0.0674925 140 After 1000 training step(s), cross entropy on all data is 0.0163385 141 After 2000 training step(s), cross entropy on all data is 0.00907547 142 After 3000 training step(s), cross entropy on all data is 0.00714436 143 After 4000 training step(s), cross entropy on all data is 0.00578471 144 145 146 w1: [[-1.96182752 2.58235407 1.68203771] 147 [-3.46817183 1.06982315 2.11788988]] 148 w2: [[-1.82471502] 149 [ 2.68546653] 150 [ 1.41819501]] 151 152 Process finished with exit code 0 153 '''
简单神经网络TensorFlow实现
学习TensorFlow笔记
|
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
|
import tensorflow as tf#定义变量#Variable 定义张量及shapew1= tf.Variable(tf.random_normal([2, 3], stddev=1, seed=1))w2= tf.Variable(tf.random_normal([3, 1], stddev=1, seed=1))with tf.Session() as sess: print(sess.run(w1.initializer)) print(sess.run(w2.initializer))#None#None#打印张量，查看数据shape等信息print(w1)print(w2)#<tf.Variable 'Variable:0' shape=(2, 3) dtype=float32_ref>#<tf.Variable 'Variable_1:0' shape=(3, 1) dtype=float32_ref>#tf.constan是一个计算，结果为一个张量，保存在变量x中x = tf.constant([[0.7, 0.9]])print(x)#Tensor("Const:0", shape=(1, 2), dtype=float32)with tf.Session() as sess: print(sess.run(x))#[[ 0.69999999 0.89999998]]#定义前向传播的神经网络#matmul做矩阵乘法a = tf.matmul(x, w1) # x shape=(1, 2) w1 shape=(2, 3)print(a)#Tensor("MatMul:0", shape=(1, 3), dtype=float32)y = tf.matmul(a, w2) #a shape=(1, 3) w2 shape=(3, 1)print(y)#Tensor("MatMul_1:0", shape=(1, 1), dtype=float32)#调用会话输出结果with tf.Session() as sess: sess.run(w1.initializer) sess.run(w2.initializer) print(sess.run(a)) #[[-2.76635647 1.12854266 0.57783246]] print(sess.run(y)) #[[ 3.95757794]]#placeholderx=tf.placeholder(tf.float32,shape=(1,2),name="input")a=tf.matmul(x,w1)y=tf.matmul(a,w2)sess=tf.Session()init_op=tf.global_variables_initializer()sess.run(init_op)print(sess.run(y,feed_dict={x:[[0.8,0.9]]}))#[[ 4.2442317]]x = tf.placeholder(tf.float32, shape=(3, 2), name="input")a = tf.matmul(x, w1)y = tf.matmul(a, w2)sess = tf.Session()#使用tf.global_variables_initializer()来初始化所有的变量init_op = tf.global_variables_initializer()sess.run(init_op)print(sess.run(y, feed_dict={x: [[0.7,0.9],[0.1,0.4],[0.5,0.8]]}))'''[[ 3.95757794] [ 1.15376544] [ 3.16749239]]''' |
整体神经网络的实现
|
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
|
import tensorflow as tffrom numpy.random import RandomState#定义神经网络的参数，输入和输出节点batch_size=8#均值为0 方差为1 随机分布满足正态分布 shape为2*3w1=tf.Variable(tf.random_normal([2,3],stddev=1,seed=1))w2=tf.Variable(tf.random_normal([3, 1], stddev=1, seed=1))#shape 根据数据自动计算 batchsize个x=tf.placeholder(tf.float32,shape=(None,2))y_=tf.placeholder(tf.float32,shape=(None,1))#定义前向传播过程，损失函数及反向传播算法a=tf.matmul(x,w1)y=tf.matmul(a,w2)#损失函数 使用交叉熵#优化方法使用AdamOptimizercross_entropy=-tf.reduce_mean(y_*tf.log(tf.clip_by_value(y,1e-10,1.0)))train_step=tf.train.AdamOptimizer(learning_rate=0.001).minimize(cross_entropy)rdm=RandomState(1)#随机生成128个数据 shape 128*2X=rdm.rand(128,2)#Y的值是模拟的 ，实际假设x2+x1如果大于1则标签Y为1 否则标签Y为0Y=[[int(x1+x2<1)] for (x1,x2) in X]#创建一个会话 ，运算计算图#全局初始化变量STEPS = 5000with tf.Session() as sess: init_op=tf.global_variables_initializer() sess.run(init_op) # 输出目前（未经训练）的参数取值。 print("w1:", sess.run(w1)) print("w2:", sess.run(w2)) print("
") for i in range(STEPS): start = (i * batch_size) % 128 end = (i * batch_size) % 128 + batch_size sess.run(train_step, feed_dict={x: X[start:end], y_: Y[start:end]}) if i % 1000 == 0: total_cross_entropy = sess.run(cross_entropy, feed_dict={x: X, y_: Y}) print("After %d training step(s), cross entropy on all data is %g" % (i, total_cross_entropy)) # 输出训练后的参数取值。 print("
") print("w1:", sess.run(w1)) print("w2:", sess.run(w2))'''w1: [[-0.81131822 1.48459876 0.06532937] [-2.4427042 0.0992484 0.59122431]]w2: [[-0.81131822] [ 1.48459876] [ 0.06532937]]After 0 training step(s), cross entropy on all data is 0.0674925After 1000 training step(s), cross entropy on all data is 0.0163385After 2000 training step(s), cross entropy on all data is 0.00907547After 3000 training step(s), cross entropy on all data is 0.00714436After 4000 training step(s), cross entropy on all data is 0.00578471w1: [[-1.96182752 2.58235407 1.68203771] [-3.46817183 1.06982315 2.11788988]]w2: [[-1.82471502] [ 2.68546653] [ 1.41819501]]Process finished with exit code 0''' |