SHOW:
|
|
- or go back to the newest paste.
| 1 | __author__ = "Bharat Kul Ratan" | |
| 2 | ||
| 3 | import random | |
| 4 | import matplotlib.pyplot as plt | |
| 5 | import seaborn as sns | |
| 6 | import pandas as pd | |
| 7 | ||
| 8 | ||
| 9 | def plot_points(): | |
| 10 | x_coordinates = pd.Series(random.sample(range(1, 20), 10), name='x_axis') | |
| 11 | y_coordinates = list(map(lambda x: 2*x+3, x_coordinates)) | |
| 12 | # Add a random value, so that y values does not lie on the equation exactly | |
| 13 | y_coordinates = pd.Series([y + random.randint(-5, 5) for y in y_coordinates], name='y_axis') | |
| 14 | ||
| 15 | x_and_y = [{'x': item[0], 'y': item[1]} for item in zip(x_coordinates, y_coordinates)] | |
| 16 | sns.regplot(x=x_coordinates, y=y_coordinates, fit_reg=False) | |
| 17 | plt.show() | |
| 18 | return x_and_y | |
| 19 | ||
| 20 | ||
| 21 | def mse(slope, intercept, points): | |
| 22 | error = 0 | |
| 23 | for item in points: | |
| 24 | predicted = item['x'] * slope + intercept | |
| 25 | output = item['y'] | |
| 26 | error += ((predicted - output) ** 2) | |
| 27 | error /= len(points) | |
| 28 | return error | |
| 29 | ||
| 30 | ||
| 31 | def update_weights(slope, intercept, learning_rate, points): | |
| 32 | m_update = 0 | |
| 33 | c_update = 0 | |
| 34 | for item in points: | |
| 35 | m_update += -2 * item['x'] * (item['y'] - (slope * item['x'] + intercept)) | |
| 36 | c_update += -2 * (item['y'] - (slope * item['x'] + intercept)) | |
| 37 | ||
| 38 | slope -= (m_update/float(len(points))) * learning_rate | |
| 39 | intercept -= (c_update/float(len(points))) * learning_rate | |
| 40 | return slope, intercept | |
| 41 | ||
| 42 | ||
| 43 | def run(): | |
| 44 | x_and_y = plot_points() | |
| 45 | check_values = 1 | |
| 46 | m = 1 | |
| 47 | c = 0 | |
| 48 | for i in range(100000): | |
| 49 | mse_error = mse(m, c, x_and_y) | |
| 50 | if i == check_values: | |
| 51 | print('slope=', round(m, 3), 'intercept=', round(c, 3), 'mse=', round(mse_error, 3)) | |
| 52 | check_values *= 10 | |
| 53 | m, c = update_weights(m, c, 0.001, x_and_y) | |
| 54 | ||
| 55 | ||
| 56 | if __name__ == '__main__': | |
| 57 | run() |
