下面用Tensorflow实现形如Y=WX+b形式的一元线性回归模型
1.数据准备及数据可视化
import numpy as np#科学计算库
import tensorflow as tf
import matplotlib.pyplot as plt
plt.rcParams["figure.figsize"] = (14,8)#作可视化时图像的长和宽
n_observations = 100#样本点数
xs = np.linspace(-3, 3, n_observations)#在-3~3之间产生一百个点
ys = np.sin(xs) + np.random.uniform(-0.5, 0.5, n_observations)#为增加随机性增加了一个随机扰动
plt.scatter(xs, ys)#作散点图
plt.show()
运行结果:
2.准备好placeholder以便于存放数据
X = tf.placeholder(tf.float32, name='X')#准备好容器存放数据
Y = tf.placeholder(tf.float32, name='Y')
3.初始化参数/权重
W = tf.Variable(tf.random_normal([1]),name='weight')#从高斯分布数据中采样
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)#依据已有的参数完成运算W,X为标量
下面计算损失函数来评估预测结果的好坏
5.计算损失函数的值
loss = tf.square(Y - Y_pred, name='loss')#这里的Y-Y_pred为向量
接下来就是要使得损失函数最小化,这里用梯度下降法,设置步长为0.01
6.定义优化器并初始化optimizer
learning_rate = 0.01#指定步长
optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(loss)#损失函数为凸函数,用梯度下降法
7.训练模型:指定迭代次数,并在session里执行graph
n_samples = xs.shape[0]
with tf.Session() as sess:
# 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])
运行结果:
Epoch 0: [ 0.76658875]
Epoch 5: [ 0.18966872]
Epoch 10: [ 0.18967324]
Epoch 15: [ 0.18967324]
Epoch 20: [ 0.18967324]
Epoch 25: [ 0.18967324]
Epoch 30: [ 0.18967324]
Epoch 35: [ 0.18967324]
Epoch 40: [ 0.18967324]
Epoch 45: [ 0.18967324]
8.最终结果可视化:
print(W,b)
plt.plot(xs, ys, 'bo', label='Real data')
plt.plot(xs, xs * W + b, 'r', label='Predicted data')
plt.legend()
plt.show()
运行结果:
[ 0.28863591] [-0.13046201]
W:0.288636
b:-0.130462