モウpython(tensorflow)まとめ-ニューラルネットワーク構築(簡易版)
2583 ワード
import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
def add_layer(inputs,in_size,out_size,activation_function=None): # , ( , , , ( ))
Weights = tf.Variable(tf.random_normal([in_size,out_size])) # ——Weights, [in_size,out_size]
biases = tf.Variable(tf.zeros([1,out_size])+0.1) # ——biases, [1,out_size] , 0.1
Wx_plus_b=tf.matmul(inputs,Weights)+biases # ,Wx_plus_b= ax+b, ,
if activation_function is None:
outputs = Wx_plus_b
else:
outputs = activation_function(Wx_plus_b)
return outputs #
##
x_data = np.linspace(-1,1,300)[:,np.newaxis] # -1 1 , 300 ,x_data [:,np.newaxis] :[1,300], 1 300 。
noise = np.random.normal(0,0.05,x_data.shape) # 0, 0.05 , x_data ( x_data )
y_data = np.square(x_data)-0.5+noise # y_data:(x_data)^2-0.5 + noise
# placeholder() graph , , 。 session, , feed_dict() 。
xs = tf.placeholder(tf.float32,[None,1]) # 1 xs
ys = tf.placeholder(tf.float32,[None,1]) # 1 ys
#
l1 = add_layer(xs,1,10,activation_function=tf.nn.relu) # xs, 1 ,10 , relu.
prediction = add_layer(l1,10,1,activation_function=None) # l1,10 ,1 , None( )
#
loss =tf.reduce_mean(tf.reduce_sum(tf.square(ys-prediction),reduction_indices=[1])) #reduction_indices //
train_step=tf.train.GradientDescentOptimizer(0.1).minimize(loss) # GDO loss
# , run , ,
init =tf.initialize_all_variables()
# sess,
sess =tf.Session()
#
sess.run(init)
# ,
fig = plt.figure()
ax = fig.add_subplot(1,1,1)
ax.scatter(x_data,y_data)
plt.ion()
plt.show()
# , 1000
for i in range(1000):
sess.run(train_step,feed_dict={xs:x_data,ys:y_data}) # , x_data y_data
if i%50==0: # 50
# print(sess.run(loss,feed_dict={xs:x_data,ys:y_data})) # 50 loss
try: # lines , , ( )
ax.lines.remove(lines[0])
except Exception:
pass
prediction_value = sess.run(prediction,feed_dict={xs:x_data}) #
lines = ax.plot(x_data,prediction_value,'r-',lw=5) # , 5 x_datahe
plt.pause(0.1) # 0.1s