全连接层的张量实现----单层神经网络的实现
定义好权值张量𝑾和偏置张量𝒃批量矩阵相乘函数 tf.matmul()即可完成网络层的计算偏置向量𝒃与计算完𝑿@𝑾的相加将结果传入激活函数
import tensorflow
as tf
x
= tf
.random
.normal
([2,567])
w1
= tf
.Variable
(tf
.random
.truncated_normal
([567,250],stddev
=0.5))
b1
= tf
.zeros
([250])
o1
= tf
.matmul
(x
,w1
)+b1
o1
= tf
.nn
.relu
(o1
)
print(o1
)
out
:
tf
.Tensor
(
[[ 0. 5.6923637 0. 0. 5.924873 6.430141
1.0681016 22.66217 2.931157 0. 0. 12.744288
0. 23.959446 15.1364975 14.34301 0. 1.439672
6.699648 3.8294516 0. 0. 0. 0.02968955
0. 29.599365 3.4065924 0. 0. 0.
16.701733 4.3234434 0. 2.4895988 24.911697 0.
1.3113308 0. 0. 17.62508 0. 0.83378553
0. 0. 0. 17.487988 10.510666 3.2711444
0. 0. 0. 0. 2.8857956 7.578166
6.4026513 0. 0. 15.8455 7.128825 2.867275
0. 11.406243 0. 0. 7.3629208 0.
0. 0. 0. 18.267078 0. 0.
0. 11.843224 18.811085 0. 0.7894125 17.675303
6.381042 0. 0. 0.90418094 0.3319425 4.561783
20.457237 0. 0. 0. 2.025149 9.29306
0. 0. 0. 1.4207561 0. 11.575405
0. 0. 0. 0. 0. 9.340631
2.9096978 2.915633 0. 4.264126 0. 0.
3.499099 10.030045 6.4256363 5.4497538 0. 0.
0. 9.498019 0. 0. 0. 3.46416
0. 1.7147312 0. 2.074488 11.75713 0.
0. 11.769547 0. 0.9206114 27.530376 1.5469017
13.124647 0.35898685 9.134467 6.418855 0. 12.321739
18.213778 5.8426123 0. 15.616054 13.56566 10.221371
0.5425372 0. 4.1035767 0. 6.5984387 18.210154
12.665659 2.4419503 3.4360933 4.824207 13.430967 0.
0. 0. 0. 18.075436 0. 0.9268236
6.993668 0. 0.28516054 2.6081998 0. 0.
0. 0. 0. 0. 4.204358 5.167027
0. 1.1935129 2.8025086 0. 0.6381178 0.
0. 0. 0. 0. 15.845655 18.041857
0. 0. 6.383477 8.501418 17.152485 0.
2.0526805 0. 0. 0. 17.907549 0.
0. 5.7903147 0. 13.597157 0. 0.
1.641685 0. 8.30833 3.2230053 0. 0.
8.802971 0. 0. 0. 0. 0.
0. 0. 0. 0. 0. 2.9737372
8.685434 17.858145 0.49902278 2.244833 0. 1.5718412
5.4440613 0. 16.20961 4.716572 16.234251 0.
7.8437977 0. 6.2740684 0. 5.991658 5.795619
0. 0. 0. 0. 8.170209 5.4254475
2.9483936 6.7834206 8.562119 0. ]
[ 0. 0. 9.827619 0. 0. 11.907226
7.934566 6.047997 10.834481 9.254445 4.593179 5.8291845
9.88936 11.618725 0. 0. 0. 6.091893
2.1074784 0. 0. 0. 0. 0.
0. 0.55824566 7.6742063 2.133997 0. 12.604022
4.1145587 14.468975 5.0262456 4.389586 0. 0.
14.072462 0. 2.8475811 5.4412966 3.3362398 9.0758915
4.297921 2.6291656 7.007786 0. 6.421891 16.863476
0. 0. 0. 0.1655035 11.46364 7.319919
9.231743 0. 0. 0.5708244 14.170641 8.481859
18.731842 9.157413 2.8565936 6.4770727 0. 0.
18.450764 1.717516 4.903287 1.5349118 0. 2.135833
25.030903 0. 22.240208 0. 0. 7.4653773
0. 0. 18.829258 3.2901583 4.663457 0.
3.6319869 12.821178 4.061035 0. 0. 0.
0. 7.68061 14.44763 0. 2.9567814 0.
6.9502516 0. 0. 0. 0. 0.
1.0752095 8.577261 0. 0. 0. 0.
15.774179 11.008202 2.6374032 8.135115 13.102028 1.7548213
0. 0. 1.7659788 0. 0.51686954 7.1937895
0. 0. 0. 0. 0. 0.
0. 5.2338986 0. 0. 0. 0.
0. 0. 8.602976 1.8720741 0. 0.5666175
0. 0. 3.6209924 2.4786031 0. 0.
2.88995 5.0474663 0. 2.6065469 18.16195 7.4969788
5.227219 9.725359 2.9107327 12.853128 0. 17.645288
12.918257 0. 0. 0. 0. 10.962885
0. 3.4822338 11.727208 0. 0. 0.9324732
4.886016 4.829459 7.7510967 13.661399 1.1002091 0.
0. 6.2028437 0. 0. 15.42531 0.73701906
0. 0. 2.587556 0. 13.945081 13.631976
25.053593 0. 0. 7.565555 13.155384 3.5761912
0. 0. 0. 0. 0. 3.1220303
0. 0. 0. 8.0910225 0. 0.
0. 0. 8.924197 0. 0. 0.
0. 0. 8.213308 1.965143 0. 11.267947
8.109399 0. 0. 2.2391272 0. 0.
0. 0. 0. 0. 0. 0.
6.3063564 0. 0. 0. 12.160679 5.7818184
0. 9.530418 15.926281 0. 6.8691816 1.3507385
15.85659 13.0173025 3.3913822 0. 0. 0.
22.77712 5.178497 0. 21.299566 ]], shape
=(2, 250), dtype
=float32
)
