poly-lr.py

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import numpy as np
import tensorflow as tf
import matplotlib
matplotlib.use('TKAgg')
from matplotlib import pyplot as plt

# 定義參數
data_size = 100
normalize_lambda = 0.02
learning_rate = 0.001
learning_loop = 500

# 產生資料
def generate_dataset():
    x_data = np.linspace(-3, 3, data_size)
    # 生成函式為 y = 0.2x^3 + 1.79x + 2
    y_data = 0.2 * x_data**3 + 1.3 * x_data + np.random.uniform(-0.5, 0.5, data_size)
    return x_data, y_data

# 定義節點圖
def multi_poly_linear_regression(power):
    x = tf.placeholder(tf.float32, shape=(None,), name="x")
    y = tf.placeholder(tf.float32, shape=(None,), name="y")
    y_pred = tf.Variable(tf.random_normal([1]), name="pred")
    loss = 0
    # 計算多項次的y_pred
    for i in range(power):
        W = tf.Variable(tf.random_normal([1]), name='weight_%d' % i)
        y_pred = tf.add(tf.multiply(tf.pow(x, i), W), y_pred)
        # 加入正規化
        loss += normalize_lambda * W ** 2
    loss += tf.reduce_sum(tf.pow(y_pred - y, 2)) / data_size
    return x, y, y_pred, loss

def run():
    x_batch, y_batch = generate_dataset()
    x, y, y_pred, loss = multi_poly_linear_regression(4)
    optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(loss)
    init = tf.global_variables_initializer()

    with tf.Session() as sess:
        sess.run(init)

        for i in range(learning_loop):
            loss_val,_ = sess.run([loss, optimizer], {x: x_batch, y: y_batch})
            y_pred_batch = sess.run(y_pred, {x:x_batch})
            
        plt.figure(1)
        plt.scatter(x_batch, y_batch)
        plt.scatter(x_batch, y_pred_batch)
        plt.show()

if __name__ == '__main__':
    run()