1、sigmoid-不使用relu
#!/usr/bin/python3
# -*- coding:utf-8 -*-
import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
"""
Logistic regression
"""
# 数据
x1 = np.random.normal(-4, 2, 1000)[:,np.newaxis] # 1000x1
x2 = np.random.normal(4,2 , 1000)[:,np.newaxis]
train_x = np.vstack((x1, x2)) # 2000x1
train_y = np.asarray([0.] * len(x1) + [1.] * len(x2))[:,np.newaxis] # 2000x1
plt.scatter(train_x, train_y)
# plt.show()
x=tf.placeholder(tf.float32,[None,1],'x')
y_=tf.placeholder(tf.float32,[None,1],'y_')
with tf.variable_scope('wb'):
w=tf.get_variable('w',(1,1),dtype=tf.float32,initializer=tf.random_uniform_initializer)
b= tf.Variable(tf.zeros([1, 1]) + 0.1)
with tf.variable_scope('wb2') as scope:
# scope.reuse_variables()
w2=tf.get_variable('w2',(10,1),dtype=tf.float32,initializer=tf.random_uniform_initializer)
b2= tf.Variable(tf.zeros([1, 1]) + 0.1)
y=tf.nn.sigmoid(tf.add(tf.matmul(x,w),b))
# y=tf.nn.relu(tf.add(tf.matmul(x,w),b))
# y=tf.nn.sigmoid(tf.add(tf.matmul(y,w2),b2))
# loss function
# loss=tf.reduce_mean(tf.reduce_sum(tf.square(y-y_),reduction_indices=[1]))
loss=tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y_,logits=y))
train_op=tf.train.GradientDescentOptimizer(0.1).minimize(loss)
sess=tf.InteractiveSession(graph=tf.get_default_graph())
tf.global_variables_initializer().run()
for step in range(1000):
sess.run(train_op,feed_dict={x:train_x,y_:train_y})
all_xs = np.linspace(-10, 10, 100)[:,np.newaxis]
prdiction_value = sess.run(y, feed_dict={x: all_xs})
lines = plt.plot(all_xs, prdiction_value, 'r-', lw=5)
plt.show()
sess.close()
结果:
2、sigmoid-使用relu
#!/usr/bin/python3
# -*- coding:utf-8 -*-
import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
"""
Logistic regression
"""
# 数据
x1 = np.random.normal(-4, 2, 1000)[:,np.newaxis] # 1000x1
x2 = np.random.normal(4,2 , 1000)[:,np.newaxis]
train_x = np.vstack((x1, x2)) # 2000x1
train_y = np.asarray([0.] * len(x1) + [1.] * len(x2))[:,np.newaxis] # 2000x1
plt.scatter(train_x, train_y)
# plt.show()
x=tf.placeholder(tf.float32,[None,1],'x')
y_=tf.placeholder(tf.float32,[None,1],'y_')
with tf.variable_scope('wb'):
w=tf.get_variable('w',(1,10),dtype=tf.float32,initializer=tf.random_uniform_initializer)
b= tf.Variable(tf.zeros([1, 10]) + 0.1)
with tf.variable_scope('wb2') as scope:
# scope.reuse_variables()
w2=tf.get_variable('w2',(10,1),dtype=tf.float32,initializer=tf.random_uniform_initializer)
b2= tf.Variable(tf.zeros([1, 1]) + 0.1)
# y=tf.nn.sigmoid(tf.add(tf.matmul(x,w),b))
y=tf.nn.relu(tf.add(tf.matmul(x,w),b))
y=tf.nn.sigmoid(tf.add(tf.matmul(y,w2),b2))
# loss function
# loss=tf.reduce_mean(tf.reduce_sum(tf.square(y-y_),reduction_indices=[1]))
loss=tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y_,logits=y))
train_op=tf.train.GradientDescentOptimizer(0.1).minimize(loss)
sess=tf.InteractiveSession(graph=tf.get_default_graph())
tf.global_variables_initializer().run()
for step in range(1000):
sess.run(train_op,feed_dict={x:train_x,y_:train_y})
all_xs = np.linspace(-10, 10, 100)[:,np.newaxis]
prdiction_value = sess.run(y, feed_dict={x: all_xs})
lines = plt.plot(all_xs, prdiction_value, 'r-', lw=5)
plt.show()
sess.close()
结果:
3、softmax(多分类)
sigmoid只能针对二分类问题,对于多分类可以使用softmax
#!/usr/bin/python3
# -*- coding:utf-8 -*-
import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
"""
Logistic regression
"""
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
# mnist图像大小是28x28 分成0~9 共10类
x=tf.placeholder(tf.float32,[None,28*28*1])
y_=tf.placeholder(tf.float32,[None,10])
with tf.variable_scope('wb'):
w=tf.get_variable('w',[28*28,10],initializer=tf.random_uniform_initializer)*0.001
b=tf.Variable(tf.zeros([10])+0.1,dtype=tf.float32)
y=tf.nn.softmax(tf.add(tf.matmul(x,w),b))
loss=tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y_,logits=y))
train_op=tf.train.AdamOptimizer(0.5).minimize(loss)
correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
# Calculate accuracy
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
sess=tf.InteractiveSession(graph=tf.get_default_graph())
tf.global_variables_initializer().run()
for step in range(1000):
batch_xs, batch_ys = mnist.train.next_batch(128)
train_op.run({x:batch_xs,y_:batch_ys})
if step % 100==0:
print("step",step,'acc',accuracy.eval({x:batch_xs,y_:batch_ys}),'loss',loss.eval({x:batch_xs,y_:batch_ys}))
# test acc
print('test acc',accuracy.eval({x:mnist.test.images,y_:mnist.test.labels}))
sess.close()
结果:
说明:
没有使用隐藏层,所以精度并不是很高,
接下来会使用DNN、CNN以及RNN来提升其精度!