1.简单线性回归
(1)数据准备
实际的数据大家可以通过pandas等package读入,也可以使用自带的Boston House Price数据集,这里为了简单,我们自己手造一点数据集。
%matplotlib inline
import numpy as np
import tensorflow as tf
import matplotlib.pyplot as plt
plt.rcParams["figure.figsize"] = (14,8)
n_observations = 100
xs = np.linspace(-3, 3, n_observations)
ys = np.sin(xs) + np.random.uniform(-0.5, 0.5, n_observations)
plt.scatter(xs, ys)
plt.show()
(2)准备好placeholder
X = tf.placeholder(tf.float32, name='X')
Y = tf.placeholder(tf.float32, name='Y')
(3)初始化参数/权重
W = tf.Variable(tf.random_normal([1]), name='weight')
b = tf.Variable(tf.random_normal([1]), name='bias')
(4)计算预测结果
Y_pred = tf.add(tf.multiply(X, W), b)
(5)计算损失函数值
loss = tf.square(Y - Y_pred, name='loss')
(6)初始化optimizer
learning_rate = 0.01
optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(loss)
(7)指定迭代次数,并在session里执行graph
n_samples = xs.shape[0]
with tf.Session() as sess:
# 记得初始化所有变量
sess.run(tf.global_variables_initializer())
writer = tf.summary.FileWriter('./graphs/linear_reg', sess.graph)
# 训练模型
for i in range(50):
total_loss = 0
for x, y in zip(xs, ys):
# 通过feed_dic把数据灌进去
_, l = sess.run([optimizer, loss], feed_dict={X: x, Y:y})
total_loss += l
if i%5 ==0:
print('Epoch {0}: {1}'.format(i, total_loss/n_samples))
# 关闭writer
writer.close()
# 取出w和b的值
W, b = sess.run([W, b])
2.多项式回归
对于非线性的数据分布,用线性回归拟合程度一般,我们来试试多项式回归
(1)数据准备
%matplotlib inline
import numpy as np
import tensorflow as tf
import matplotlib.pyplot as plt
plt.rcParams["figure.figsize"] = (14,8)
n_observations = 100
xs = np.linspace(-3, 3, n_observations)
ys = np.sin(xs) + np.random.uniform(-0.5, 0.5, n_observations)
plt.scatter(xs, ys)
plt.show()
(2)准备好placeholder
X = tf.placeholder(tf.float32, name='X')
Y = tf.placeholder(tf.float32, name='Y')
(3)初始化参数/权重
W = tf.Variable(tf.random_normal([1]), name='weight')
b = tf.Variable(tf.random_normal([1]), name='bias')
(4)计算预测结果
Y_pred = tf.add(tf.multiply(X, W), b)
#添加高次项
W_2 = tf.Variable(tf.random_normal([1]), name='weight_2')
Y_pred = tf.add(tf.multiply(tf.pow(X, 2), W_2), Y_pred)
W_3 = tf.Variable(tf.random_normal([1]), name='weight_3')
Y_pred = tf.add(tf.multiply(tf.pow(X, 3), W_3), Y_pred)
(5)计算损失函数值
sample_num = xs.shape[0]
loss = tf.reduce_sum(tf.pow(Y_pred - Y, 2)) / sample_num
(6)初始化optimizer
learning_rate = 0.01
optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(loss)
(7)指定迭代次数,并在session里执行graph
n_samples = xs.shape[0]
with tf.Session() as sess:
# 记得初始化所有变量
sess.run(tf.global_variables_initializer())
writer = tf.summary.FileWriter('./graphs/polynomial_reg', sess.graph)
# 训练模型
for i in range(1000):
total_loss = 0
for x, y in zip(xs, ys):
# 通过feed_dic把数据灌进去
_, l = sess.run([optimizer, loss], feed_dict={X: x, Y:y})
total_loss += l
if i%20 ==0:
print('Epoch {0}: {1}'.format(i, total_loss/n_samples))
# 关闭writer
writer.close()
# 取出w和b的值
W, W_2, W_3, b = sess.run([W, W_2, W_3, b])
print("W:"+str(W[0]))
print("W_2:"+str(W_2[0]))
print("W_3:"+str(W_3[0]))
print("b:"+str(b[0]))
plt.plot(xs, ys, 'bo', label='Real data')
plt.plot(xs, xs*W + np.power(xs,2)*W_2 + np.power(xs,3)*W_3 + b, 'r', label='Predicted data')
plt.legend()
plt.show()
3.逻辑回归
(1)数据准备
import os
os.environ['TF_CPP_MIN_LOG_LEVEL']='2'
import numpy as np
import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
import time
#使用tensorflow自带的工具加载MNIST手写数字集合
mnist = input_data.read_data_sets('/data/mnist', one_hot=True)
#查看一下数据维度
mnist.train.images.shape
#查看target维度
mnist.train.labels.shape
(2)准备好placeholder
batch_size = 128
X = tf.placeholder(tf.float32, [batch_size, 784], name='X_placeholder')
Y = tf.placeholder(tf.int32, [batch_size, 10], name='Y_placeholder')
(3)初始化参数/权重
w = tf.Variable(tf.random_normal(shape=[784, 10], stddev=0.01), name='weights')
b = tf.Variable(tf.zeros([1, 10]), name="bias")
(4)拿到每个类别的score
logits = tf.matmul(X, w) + b
(5)计算多分类softmax的loss function
# 求交叉熵损失
entropy = tf.nn.softmax_cross_entropy_with_logits(logits=logits, labels=Y, name='loss')
# 求平均
loss = tf.reduce_mean(entropy)
(6)初始化optimizer
这里的最优化用的是随机梯度下降,我们可以选择AdamOptimizer这样的优化器
learning_rate = 0.01
optimizer = tf.train.AdamOptimizer(learning_rate).minimize(loss)
(7)指定迭代次数,并在session里执行graph
#迭代总轮次
n_epochs = 30
with tf.Session() as sess:
# 在Tensorboard里可以看到图的结构
writer = tf.summary.FileWriter('./graphs/logistic_reg', sess.graph)
start_time = time.time()
sess.run(tf.global_variables_initializer())
n_batches = int(mnist.train.num_examples/batch_size)
for i in range(n_epochs): # 迭代这么多轮
total_loss = 0
for _ in range(n_batches):
X_batch, Y_batch = mnist.train.next_batch(batch_size)
_, loss_batch = sess.run([optimizer, loss], feed_dict={X: X_batch, Y:Y_batch})
total_loss += loss_batch
print('Average loss epoch {0}: {1}'.format(i, total_loss/n_batches))
print('Total time: {0} seconds'.format(time.time() - start_time))
print('Optimization Finished!')
# 测试模型
preds = tf.nn.softmax(logits)
correct_preds = tf.equal(tf.argmax(preds, 1), tf.argmax(Y, 1))
accuracy = tf.reduce_sum(tf.cast(correct_preds, tf.float32))
n_batches = int(mnist.test.num_examples/batch_size)
total_correct_preds = 0
for i in range(n_batches):
X_batch, Y_batch = mnist.test.next_batch(batch_size)
accuracy_batch = sess.run([accuracy], feed_dict={X: X_batch, Y:Y_batch})
total_correct_preds += accuracy_batch[0]
print('Accuracy {0}'.format(total_correct_preds/mnist.test.num_examples))
writer.close()
4.多层感知机
(1)数据准备
import numpy as np
import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
import time
#使用tensorflow自带的工具加载MNIST手写数字集合
mnist = input_data.read_data_sets('/data/mnist', one_hot=True)
#查看一下数据维度
mnist.train.
(2)准备好placeholder
X = tf.placeholder(tf.float32, [None, 784], name='X_placeholder')
Y = tf.placeholder(tf.int32, [None, 10], name='Y_placeholder')
(3)初始化参数/权重
# 网络参数
n_hidden_1 = 256 # 第1个隐层
n_hidden_2 = 256 # 第2个隐层
n_input = 784 # MNIST 数据输入(28*28*1=784)
n_classes = 10 # MNIST 总共10个手写数字类别
weights = {
'h1': tf.Variable(tf.random_normal([n_input, n_hidden_1]), name='W1'),
'h2': tf.Variable(tf.random_normal([n_hidden_1, n_hidden_2]), name='W2'),
'out': tf.Variable(tf.random_normal([n_hidden_2, n_classes]), name='W')
}
biases = {
'b1': tf.Variable(tf.random_normal([n_hidden_1]), name='b1'),
'b2': tf.Variable(tf.random_normal([n_hidden_2]), name='b2'),
'out': tf.Variable(tf.random_normal([n_classes]), name='bias')
}
(4)构建网格计算graph
def multilayer_perceptron(x, weights, biases):
# 第1个隐层,使用relu激活函数
layer_1 = tf.add(tf.matmul(x, weights['h1']), biases['b1'], name='fc_1')
layer_1 = tf.nn.relu(layer_1, name='relu_1')
# 第2个隐层,使用relu激活函数
layer_2 = tf.add(tf.matmul(layer_1, weights['h2']), biases['b2'], name='fc_2')
layer_2 = tf.nn.relu(layer_2, name='relu_2')
# 输出层
out_layer = tf.add(tf.matmul(layer_2, weights['out']), biases['out'], name='fc_3')
return out_layer
(5)拿到预测类别score
pred = multilayer_perceptron(X, weights, biases)
(6)计算损失函数值并初始化optimizer
learning_rate = 0.001
loss_all = tf.nn.softmax_cross_entropy_with_logits(logits=pred, labels=Y, name='cross_entropy_loss')
loss = tf.reduce_mean(loss_all, name='avg_loss')
optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(loss)
(7)初始化变量
init = tf.global_variables_initializer()
(8)在session中执行graph定义的运算
#训练总轮数
training_epochs = 15
#一批数据大小
batch_size = 128
#信息展示的频度
display_step = 1
with tf.Session() as sess:
sess.run(init)
writer = tf.summary.FileWriter('./graphs/MLP_DNN', sess.graph)
# 训练
for epoch in range(training_epochs):
avg_loss = 0.
total_batch = int(mnist.train.num_examples/batch_size)
# 遍历所有的batches
for i in range(total_batch):
batch_x, batch_y = mnist.train.next_batch(batch_size)
# 使用optimizer进行优化
_, l = sess.run([optimizer, loss], feed_dict={X: batch_x, Y: batch_y})
# 求平均的损失
avg_loss += l / total_batch
# 每一步都展示信息
if epoch % display_step == 0:
print("Epoch:", '%04d' % (epoch+1), "cost=", \
"{:.9f}".format(avg_loss))
print("Optimization Finished!")
# 在测试集上评估
correct_prediction = tf.equal(tf.argmax(pred, 1), tf.argmax(Y, 1))
# 计算准确率
accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))
print("Accuracy:", accuracy.eval({X: mnist.test.images, Y: mnist.test.labels}))
writer.close()