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


def add_layer(inputs, in_size, out_size, activation_function=None):  # activation_function=None线性函数
    Weights = tf.Variable(tf.random_normal([in_size, out_size]))  # Weight中都是随机变量
    biases = tf.Variable(tf.zeros([1, out_size]) + 0.1)  # biases推荐初始值不为0
    Wx_plus_b = tf.matmul(inputs, Weights) + biases  # inputs*Weight+biases
    if activation_function is None:
        outputs = Wx_plus_b
    else:
        outputs = activation_function(Wx_plus_b)
    return outputs


# 创建数据x_data，y_data
x_data = np.linspace(-1, 1, 300)[:, np.newaxis]  # [-1,1]区间，300个单位，np.newaxis增加维度
noise = np.random.normal(0, 0.05, x_data.shape)  # 噪点
y_data = np.square(x_data) - 0.5 + noise

xs = tf.placeholder(tf.float32, [None, 1])
ys = tf.placeholder(tf.float32, [None, 1])
# 三层神经，输入层（1个神经元），隐藏层（10神经元），输出层（1个神经元）
l1 = add_layer(xs, 1, 10, activation_function=tf.nn.relu)  # 输入层
prediction = add_layer(l1, 10, 1, activation_function=None)  # 隐藏层

# predition值与y_data差别
loss = tf.reduce_mean(
    tf.reduce_sum(tf.square(ys - prediction), reduction_indices=[1]))  # square()平方,sum()求和,mean()平均值

train_step = tf.train.GradientDescentOptimizer(0.01).minimize(loss)  # 反向传播，0.1学习效率,minimize(loss)减小loss误差

init = tf.initialize_all_variables()
sess = tf.Session()
sess.run(init)  # 先执行init

# 可视化
fig = plt.figure()
ax = fig.add_subplot(1, 1, 1)
ax.scatter(x_data, y_data)
plt.ion()  # 不让show() block
plt.show()

# 训练1k次
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])  # lines建一个抹除一个
        except Exception:
            pass
            # print(sess.run(loss,feed_dict={xs:x_data,ys:y_data})) #输出loss值
        # 可视化
        prediction_value = sess.run(prediction, feed_dict={xs: x_data, ys: y_data})
        lines = ax.plot(x_data, prediction_value, 'r-', lw=5)  # x_data X轴，prediction_value Y轴，'r-'红线，lw=5线宽5
        plt.pause(0.1)  # 暂停0.1秒