人工智能TensorFlow-非线性回归

import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt

# 生成两百个随机点
x_data = np.linspace(-0.5, 0.5, 200)[:,np.newaxis]
# 生成与x_data形状一样的干扰值
noise = np.random.normal(0, 0.02, x_data.shape)
# 定义一个函数
y_data = np.square(x_data) + noise

# 定义两个placeholder
x = tf.placeholder(tf.float32, [None, 1])
y = tf.placeholder(tf.float32, [None, 1])

# 构建神经网络中间层
Weights_L1 = tf.Variable(tf.random_normal([1, 10]))
biases_L1 = tf.Variable(tf.zeros([1, 10]))
Wx_plus_b_L1 = tf.matmul(x, Weights_L1) + biases_L1
L1 = tf.nn.tanh(Wx_plus_b_L1)

# 定义输出层
Weights_L2 = tf.Variable(tf.random_normal([10, 1])) 
biases_L2 = tf.Variable(tf.zeros([1, 1]))
Wx_plus_b_L2 = tf.matmul(L1, Weights_L2) + biases_L2
prediction = tf.nn.tanh(Wx_plus_b_L2)

# 二次代价函数
loss = tf.reduce_mean(tf.square(y-prediction))
# 使用梯度下降法
train_step = tf.train.GradientDescentOptimizer(0.1).minimize(loss)

with tf.Session() as sess:
    # 变量的初始化 
    sess.run(tf.global_variables_initializer())
    # 开始训练
    for _ in range(2000):
        sess.run(train_step, feed_dict={x:x_data, y:y_data})
    # 获得预测值
    prediction_value = sess.run(prediction, feed_dict={x:x_data})
    # 画图
    plt.figure()
    plt.scatter(x_data, y_data)
    plt.plot(x_data, prediction_value, 'r-', lw=5)
    plt.show()

np.linspace(-0.5, 0.5, 200) 指在-5到5的范围里，生成200个随机数

[:,np.newaxis]将所得一维数据再添加一个维度，形成二维数据