tensorflow学习笔记(5) - 激励函数的使用

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


def add_layer(inputs, in_size, out_size, activate_function=None):
    Weights = tf.Variable(tf.random_normal(
        [in_size, out_size]))  # 有in_size行out_size列的矩阵
    biases = tf.Variable(tf.zeros([1, out_size])+0.1)  # 1行 out_size列
    Wx_plus_b = tf.matmul(inputs, Weights)+biases
    '''
    inputs: 1*in_size
    Weights: in_size*out_size
    Wx_plus_b: 1*out_size
    '''
    if activate_function is None:
        outputs = Wx_plus_b
    else:
        outputs = activate_function(Wx_plus_b)
    return outputs


x_data = np.linspace(-1, 1, 300)[:, None]  # 300行-1到1的等差数列
print(x_data)
noise = np.random.normal(0, 0.05, x_data.shape)  # 噪点方差为0.05，最小是0
y_data = np.square(x_data)-0.5+noise

xs = tf.placeholder(tf.float32, [None, 1])
ys = tf.placeholder(tf.float32, [None, 1])  # None表示的是给多少例子都可以

l1 = add_layer(xs, 1, 10, activate_function=tf.nn.relu)
prediction = add_layer(l1, 10, 1, activate_function=None)

loss = tf.reduce_mean(tf.reduce_sum(tf.square(ys-prediction),
                                    reduction_indices=[1]))
# reduction_indices=[1]：按行求和
# reduction_indices=[0]：按列求和
train_step = tf.train.GradientDescentOptimizer(0.1).minimize(loss)  # 学习率为0.1
init = tf.initialize_all_variables()
with tf.Session() as sess:
    sess.run(init)
    fig = plt.figure()
    ax = fig.add_subplot(1, 1, 1)
    ax.scatter(x_data, y_data)
    plt.ion()
    plt.show()

    for i in range(1000):
        sess.run(train_step, feed_dict={xs: x_data, ys: y_data})
        if i % 50 == 0:
            try:
                ax.lines.remove(lines[0])  # 擦除
            except Exception:
                pass

            print(sess.run(loss, feed_dict={xs: x_data, ys: y_data}))
            prediction_value = sess.run(prediction, feed_dict={
                                        xs: x_data})
            lines = ax.plot(x_data, prediction_value, 'r-', lw=5)
            plt.pause(0.2)
    print('Press any key to exit...')
    input()

结果: