通过利用TensorFlow来构建一个单层的回归预测模型,使用梯度下降算法来优化损失函数,数据是通过模拟的一个线性的数据集。
import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
if __name__ == "__main__":
#设置参数
#设置梯度下降算法的学习率
learning_rate = 0.01
#设置迭代次数
max_steps = 1000
#每迭代100次输出一次loss
show_step = 100
# 模拟训练数据
train_X = np.asarray([3.3, 4.4, 5.5, 6.71, 6.93, 4.168, 9.779, 6.182, 7.59, 2.167,
7.042, 10.791, 5.313, 7.997, 5.654, 9.27, 3.1])
train_Y = np.asarray([1.7, 2.76, 2.09, 3.19, 1.694, 1.573, 3.366, 2.596, 2.53, 1.221,
2.827, 3.465, 1.65, 2.904, 2.42, 2.94, 1.3])
#获取训练数据的大小
n_samples = train_X.shape[0]
#定义输入变量,变量只有一个特征
X = tf.placeholder(dtype=tf.float32)
#定义输出变量,输出只有一个值
Y = tf.placeholder(dtype=tf.float32)
#设计模型
#定义权重
rng = np.random
W = tf.Variable(rng.randn(),name="weight")
#定义偏置
b = tf.Variable(rng.randn(),name="bias")
#计算预测值
pred = tf.add(tf.multiply(X,W),b)
#定义损失函数
loss = 0.5 * tf.reduce_sum(tf.pow(pred-Y,2)) / n_samples
#使用梯度下降算法来优化损失函数
train_step = tf.train.GradientDescentOptimizer(learning_rate).minimize(loss)
#创建一个会话
with tf.Session() as sess:
# 初始化所有变量
initialize = tf.global_variables_initializer().run()
#迭代训练
for step in range(max_steps):
for (x,y) in zip(train_X,train_Y):
sess.run(train_step,feed_dict={X:x,Y:y})
if step % show_step == 0:
#计算模型在数据集上的损失
step_loss = sess.run([loss],feed_dict={X:train_X,Y:train_Y})
print("step:",step,"-step loss:%.4f",step_loss)
#计算最终的Loss
train_loss = sess.run([loss],feed_dict={X:train_X,Y:train_Y})
print("train loss:%.4f",train_loss)
#输出参数
print("weights:",sess.run(W),"-bias:",sess.run(b))
plt.plot(train_X,train_Y,"ro",label="original data")
plt.plot(train_X,sess.run(pred,feed_dict={X:train_X}),label="predict data")
plt.legend(loc="upper left")
plt.show()
#测试数据
test_X = np.asarray([6.83, 4.668, 8.9, 7.91, 5.7, 8.7, 3.1, 2.1])
test_Y = np.asarray([1.84, 2.273, 3.2, 2.831, 2.92, 3.24, 1.35, 1.03])
#计算回归模型在测试数据上的loss
print("test loss:%.4f",sess.run(loss,feed_dict={X:test_X,Y:test_Y}))
#绘图
plt.plot(test_X,test_Y,"bo",label="test data")
plt.plot(test_X,sess.run(pred,feed_dict={X:test_X}),label="predict test")
plt.legend(loc="upper left")
plt.show()