神经网络之数据拟合

近期学习神经网络，初步实现利用神经网络对训练集进行拟合。
import tensorflow as tf
import numpy as np
import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
import matplotlib.pyplot as plt

#定义在(-0.5,0.5)之间随机生成50个数,并转换为50*1矩阵
x_data = np.linspace(-0.5,0.5,50)[:,np.newaxis]
#随机生成50个噪音值
noise = np.random.normal(0,0.02,x_data.shape)
#y=x^2+noise
y_data = np.square(x_data) + noise

# x=tf.placeholder(tf.float32,[None,1])
x=tf.compat.v1.placeholder(tf.float32,[None,1])
# y=tf.placeholder(tf.float32,[None,1])
y=tf.compat.v1.placeholder(tf.float32,[None,1])

w_L1=tf.Variable(tf.random.normal([1,10]))
b_L1=tf.Variable(tf.zeros([1,10]))
wx_plut_b_L1=tf.matmul(x,w_L1)+b_L1
L1 = tf.nn.sigmoid(wx_plut_b_L1)

w_L2=tf.Variable(tf.random.normal([10,1]))
b_L2=tf.Variable(tf.zeros([1,1]))
wx_plut_b_L2=tf.matmul(L1,w_L2)+b_L2
prediction= tf.nn.sigmoid(wx_plut_b_L2)

loss = tf.reduce_mean(tf.square(y-prediction))
train_step=tf.compat.v1.train.ProximalGradientDescentOptimizer(0.2).minimize(loss)

with tf.compat.v1.Session() as sess:
    sess.run(tf.compat.v1.global_variables_initializer())
    for i in range(1000):
        sess.run(train_step, feed_dict={x:x_data,y:y_data})
        # percent = 1.0 * i / 1000 * 100
        # print('complete percent:%10.8s%s' % (str(percent), '%'), end='\r')
        # time.sleep(0.001)
        if i % 100 == 0:
            print(i,sess.run(loss,feed_dict={x:x_data,y:y_data}))
    prediction_value=sess.run(prediction,feed_dict={x:x_data})

    plt.figure()
    plt.scatter(x_data,y_data)
    plt.plot(x_data,prediction_value,'r-',lw=5)
    plt.show()

通过该表步长和训练次数，得到不通拟合结果，当训练次数达到10万次时，结果误差到达0.0004附近，可以看到结果已趋近于理想，也可以通过改变步长实现不同步长下训练效果对比