gradient_descent.py

import csv


def gradient_desc(m, b, data, learning_rate):
  """Gradient Descent"""
  N = len(data)
  for i in range(N):
    x = float(data[i][0])
    y = float(data[i][1])
    error = y - (m * x + b)
    m += 2 * error * x * learning_rate
    b += 2 * error * learning_rate
  return m, b


def gradient_desc_runner(data, m, b, learning_rate, iteration):
  """Run gradient_desc iteration times"""
  for i in range(iteration):
    m, b = gradient_desc(m, b, data, learning_rate)
  return m, b


def total_error(m, b, data):
  """Calculate total error of data"""
  total_error = 0
  for i in range(len(data)):
      x = float(data[i][0])
      y = float(data[i][1])
      total_error += (y - (m * x + b)) ** 2
  return total_error / float(len(data))


def run():
  """Run the program"""
  # read the dataset
  with open('data.csv', newline='') as csvfile:
    data = list(csv.reader(csvfile))

  iteration = 100000
  learning_rate = 0.0002
  m = 0  # inital m
  b = 0  # inital b
  print(f"Starting gradient decent at m = {m}, b = {b}, error={total_error(m, b, data)}")
  print("[*] Started Running Algorithm....")
  m, b = gradient_desc_runner(data, m, b, learning_rate, iteration)
  print(f"After running {iteration} iteration m = {m}, b = {b}, error={total_error(m, b, data)}")

if __name__ == "__main__":
  run()