【Python】scikit-learn机器学习（二）——BP神经网络

矫情吗；* 2022-06-09 14:48 274阅读 0赞

## 环境 ##

Mac os

python3.6

## 代码 ##

# coding:utf-8
    
    """
    Author: roguesir
    Date: 2017/8/30
    GitHub: https://roguesir.github.com
    Blog: http://blog.csdn.net/roguesir
    """
    
    from __future__ import print_function
    import tensorflow as tf
    import numpy as np
    
    def add_layer(inputs, in_size, out_size, activation_function=None):
        # add one more layer and return the output of this layer
        Weights = tf.Variable(tf.random_normal([in_size, out_size]))
        biases = tf.Variable(tf.zeros([1, out_size]) + 0.1)
        Wx_plus_b = tf.matmul(inputs, Weights) + biases
        if activation_function is None:
            outputs = Wx_plus_b
        else:
            outputs = activation_function(Wx_plus_b)
        return outputs
    
    # Make up some real data
    x_data = np.linspace(-1,1,300)[:, np.newaxis]
    noise = np.random.normal(0, 0.05, x_data.shape)
    y_data = np.square(x_data) - 0.5 + noise
    
    # define placeholder for inputs to network
    xs = tf.placeholder(tf.float32, [None, 1])
    ys = tf.placeholder(tf.float32, [None, 1])
    # add hidden layer
    l1 = add_layer(xs, 1, 10, activation_function=tf.nn.relu)
    # add output layer
    prediction = add_layer(l1, 10, 1, activation_function=None)
    
    # the error between prediction and real data
    loss = tf.reduce_mean(tf.reduce_sum(tf.square(ys - prediction),
                         reduction_indices=[1]))
    train_step = tf.train.GradientDescentOptimizer(0.1).minimize(loss)
    
    # important step
    if int((tf.__version__).split('.')[1]) < 12:
        init = tf.initialize_all_variables()
    else:
        init = tf.global_variables_initializer()
    
    sess = tf.Session()
    sess.run(init)
    
    for i in range(1000):
        # training
        sess.run(train_step, feed_dict={xs: x_data, ys: y_data})
        if i % 50 == 0:
            # to see the step improvement
            print(sess.run(loss, feed_dict={xs: x_data, ys: y_data}))