Skip to content
Related Articles

Related Articles

Gradient Descent in Linear Regression

Improve Article
Save Article
  • Difficulty Level : Easy
  • Last Updated : 23 Aug, 2022
Improve Article
Save Article

In linear regression, the model targets to get the best-fit regression line to predict the value of y based on the given input value (x). While training the model, the model calculates the cost function which measures the Root Mean Squared error between the predicted value (pred) and true value (y). The model targets to minimize the cost function. 
To minimize the cost function, the model needs to have the best value of θ1 and θ2. Initially model selects θ1 and θ2 values randomly and then iteratively update these value in order to minimize the cost function until it reaches the minimum. By the time model achieves the minimum cost function, it will have the best θ1 and θ2 values. Using these finally updated values of θ1 and θ2 in the hypothesis equation of linear equation, the model predicts the value of x in the best manner it can. 
Therefore, the question arises – How do θ1 and θ2 values get updated? 
Linear Regression Cost Function:
 

 

Gradient Descent Algorithm For Linear Regression 
 

 

 

-> θj     : Weights of the hypothesis.
-> hθ(xi) : predicted y value for ith input.
-> j     : Feature index number (can be 0, 1, 2, ......, n).
-> α     : Learning Rate of Gradient Descent.

Machine-Learning-Course

We graph cost function as a function of parameter estimates i.e. parameter range of our hypothesis function and the cost resulting from selecting a particular set of parameters. We move downward towards pits in the graph, to find the minimum value. The way to do this is taking derivative of cost function as explained in the above figure. Gradient Descent step-downs the cost function in the direction of the steepest descent. The size of each step is determined by parameter α known as Learning Rate
In the Gradient Descent algorithm, one can infer two points : 
 

  • If slope is +ve : θj = θj – (+ve value). Hence value of θj decreases.

  • If slope is -ve : θj = θj – (-ve value). Hence value of θj increases.

The choice of correct learning rate is very important as it ensures that Gradient Descent converges in a reasonable time. : 
 

  • If we choose α to be very large, Gradient Descent can overshoot the minimum. It may fail to converge or even diverge. 
     

  • If we choose α to be very small, Gradient Descent will take small steps to reach local minima and will take a longer time to reach minima. 
     

For linear regression Cost, the Function graph is always convex shaped.

Python3




# Implementation of gradient descent in linear regression 
import numpy as np 
import matplotlib.pyplot as plt 
  
class Linear_Regression: 
    def __init__(self, X, Y): 
        self.X =
        self.Y =
        self.b = [0, 0
      
    def update_coeffs(self, learning_rate): 
        Y_pred = self.predict() 
        Y = self.Y 
        m = len(Y) 
        self.b[0] = self.b[0] - (learning_rate * ((1/m) *
                                np.sum(Y_pred - Y))) 
  
        self.b[1] = self.b[1] - (learning_rate * ((1/m) *
                                np.sum((Y_pred - Y) * self.X))) 
  
    def predict(self, X=[]): 
        Y_pred = np.array([]) 
        if not X: X = self.X 
        b = self.b 
        for x in X: 
            Y_pred = np.append(Y_pred, b[0] + (b[1] * x)) 
  
        return Y_pred 
      
    def get_current_accuracy(self, Y_pred): 
        p, e = Y_pred, self.Y 
        n = len(Y_pred) 
        return 1-sum
            
                abs(p[i]-e[i])/e[i] 
                for i in range(n) 
                if e[i] != 0
        )/
    #def predict(self, b, yi): 
  
    def compute_cost(self, Y_pred): 
        m = len(self.Y) 
        J = (1 / 2*m) * (np.sum(Y_pred - self.Y)**2
        return
  
    def plot_best_fit(self, Y_pred, fig): 
                f = plt.figure(fig) 
                plt.scatter(self.X, self.Y, color='b'
                plt.plot(self.X, Y_pred, color='g'
                f.show() 
  
  
def main(): 
    X = np.array([i for i in range(11)]) 
    Y = np.array([2*i for i in range(11)]) 
  
    regressor = Linear_Regression(X, Y) 
  
    iterations = 0
    steps = 100
    learning_rate = 0.01
    costs = [] 
      
    #original best-fit line 
    Y_pred = regressor.predict() 
    regressor.plot_best_fit(Y_pred, 'Initial Best Fit Line'
      
  
    while 1
        Y_pred = regressor.predict() 
        cost = regressor.compute_cost(Y_pred) 
        costs.append(cost) 
        regressor.update_coeffs(learning_rate) 
          
        iterations += 1
        if iterations % steps == 0
            print(iterations, "epochs elapsed"
            print("Current accuracy is :"
                regressor.get_current_accuracy(Y_pred)) 
  
            stop = input("Do you want to stop (y/*)??"
            if stop == "y"
                break
  
    #final best-fit line 
    regressor.plot_best_fit(Y_pred, 'Final Best Fit Line'
  
    #plot to verify cost function decreases 
    h = plt.figure('Verification'
    plt.plot(range(iterations), costs, color='b'
    h.show() 
  
    # if user wants to predict using the regressor: 
    regressor.predict([i for i in range(10)]) 
  
if __name__ == '__main__'
    main() 


Output:

Note: Gradient descent sometimes is also implemented using Regularization.


My Personal Notes arrow_drop_up
Related Articles

Start Your Coding Journey Now!