Gradient_descent_Method

#Python Program to illustrate the method of Gradient Descent algorithm(which finds the coefficients and intercept for the best fit line or Regeression function)

import numpy as np

def gradient_descent(x,y):
    m_curr = 0
    b_curr = 0
    n = len(x)
    lr = 0.03
    iteration = 1000
    for i in range(iteration):
        y_pred = m_curr * x + b_curr

        #cost = (1/n)*sum([val**2 for val in (y-y_pred)])   //Alt method to calculate cost function

        s = 0
        for j in range(len(y)):
            s = s + (y[j] - y_pred[j])**2
        cost = (1 / n)*s

        m_der = -(2 / n) *sum(x*(y - y_pred))
        b_der = -(2 / n) *sum(y-y_pred)
        m_curr = m_curr-lr*m_der
        b_curr = b_curr-lr*b_der
        print("iteration : %.3f" %i, "\tm : %.3f" %m_curr, "\tb : %.3f" %b_curr, "\tcost func : %.3f" %cost)

x = np.array([1,2,3,4,5])
y = np.array([5,7,9,11,13])

gradient_descent(x,y)