Open in App
Not now

• Difficulty Level : Easy
• Last Updated : 01 Mar, 2023

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 Î¸1 and Î¸2. Initially model selects Î¸1 and Î¸2 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 Î¸1 and Î¸2 values. Using these finally updated values of Î¸1 and Î¸2 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 Î¸1 and Î¸2 values get updated?
Linear Regression Cost Function:

Gradient Descent Algorithm For Linear Regression

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

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 Î± known as Learning Rate
In the Gradient Descent algorithm, one can infer two points :

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

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

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

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

• If we choose Î± 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 ``=` `X ` `        ``self``.Y ``=` `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``] ` `        ``)``/``n ` `    ``#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` `J `   `    ``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.

Gradient Descent is a popular optimization algorithm for linear regression models that involves iteratively adjusting the model parameters to minimize the cost function. Here are some advantages and disadvantages of using Gradient Descent for linear regression:

• Flexibility: Gradient Descent can be used with various cost functions and can handle non-linear regression problems.
• Scalability: Gradient Descent is scalable to large datasets since it updates the parameters for each training example one at a time.
• Convergence: Gradient Descent can converge to the global minimum of the cost function, provided that the learning rate is set appropriately.