梳理tensorflow代码1——简单概念

tensorflow 主要由计算图Graph、张量tensor、会话Session组成

import tensorflow as tf
a=tf.constant([1.0,2.0],name='a')
b=tf.constant([2.0,3.0],name='b')
result=a+b  #result为张量tensor，亦可以称为节点
print(a.graph is tf.get_default_graph())
print(result)  #张量的命名，node：source_output形式
with tf.Session() as sess:
    print(sess.run(result))

output：
True
Tensor(“add:0”, shape=(2,), dtype=float32)
[ 3. 5.]

sess=tf.Session()
#通过对比上下两种会话方式，可以略知一二。
print(result.eval(session=sess))
sess.close()

[ 3.  5.]

g1=tf.Graph()
with g1.as_default():
    v=tf.get_variable('v',initializer=tf.zeros_initializer()(shape=[4]))
g2=tf.Graph()
with g2.as_default():
    v=tf.get_variable('v',initializer=tf.ones_initializer()(shape=[1]))   
with tf.Session(graph=g1) as sess:
    tf.global_variables_initializer().run()
    with tf.variable_scope('',reuse=True):
        print(sess.run(tf.get_variable('v')))
#initializer=tf.zeros_initializer要在背后加上（）

[ 0.  0.  0.  0.]

sess=tf.InteractiveSession()#将节点所在的会话设置为默认会话
config=ConfigProto(allow_sof_placement=True,log_device_placement=True)

g1=tf.Graph()
with g1.as_default():
    w1=tf.Variable(tf.random_normal([2,3],stddev=1))
    w2=tf.Variable(tf.random_normal([3,1],stddev=1))
    x=tf.placeholder(tf.float32,shape=[1,2],name='input')
    a=tf.matmul(x,w1)
    y=tf.matmul(a,w2)
with tf.Session(graph=g1) as sess:
    tf.global_variables_initializer().run()
    print(sess.run(y,feed_dict={x:[[0.7,0.9]]}))

[[ 0.49852759]]

tf.__version__

'1.4.0'

v1=tf.constant([1.0,2.0,3.0,4.0])
v2=tf.constant([4.0,3.0,2.0,1.0])
sess=tf.InteractiveSession()
a=sess.run(tf.greater(v1,v2))
sess.close()
print(a)

Output：
variables
[False False True True]

#配置会话（并行线程数、GPU分配策略、运算超时时间等）
config=tf.ConfigProto(allow_soft_placement=True,log_device_placement=True)
sess1=tf.Session(config=config)
#allow_soft_placement有GPU下一般设为True
#log_device_placement为True时，日志中将会记录每个节点被安排在哪个设备上以方便调试；生产环境中设置为False，
#减少日志量

#简易的初始化权重函数
def get_weight(shape,lamda):
    var=tf.Variable(tf.random_normal(shape),dtype=tf.float32)
    tf.add_to_collection('losses',
                        tf.contrib.layers.l2_regularizer(0.1)(var))
    return var
#每一层网络中节点的个数
layer_dimension=[2,10,10,10,1]
#网络层数
n_layers=len(layer_dimension)
x=tf.placeholder(tf.float32,shape=(None,2))
y_=tf.placeholder(tf.float32,shape=(None,1))
batch_size=8
cur_layer=x
in_dimension=layer_dimension[0]
for i in range(1,n_layers):
    out_dimension=layer_dimension[i]
    weight=get_weight([in_dimension,out_dimension],0.001)
    bias=tf.Variable(tf.constant(0.1,shape=[out_dimension]))
    cur_layer=tf.nn.relu(tf.matmul(cur_layer,weight)+bias)
    in_dimension=layer_dimension[i]
mse_loss=tf.reduce_mean(tf.square(y_-cur_layer))
tf.add_to_collection('losses',mse_loss)
loss=tf.add_n(tf.get_collection('losses'))
#通过设定get_weight函数来初始化权重
#简易地构造激活函数为relu的全连接层神经网络