Scikit - Linear Regression - Extracting Metrics

Question

Python 3
scikit-learn-0.23.1
numpy-1.18.4
scipy-1.4.1

Is it possible to extract the number of iterations it took to fit the regression model (by Gradient Descent)? Aside from the general model evaluation metrics, I don't see anything related to iteration number.

regressor = LinearRegression()  
regressor.fit(xtr, ytr)

print('Mean Absolute Error:', metrics.mean_absolute_error(yt, y_pred))  
print('Mean Squared Error:', metrics.mean_squared_error(yt, y_pred))  
print('Root Mean Squared Error:', np.sqrt(metrics.mean_squared_error(yt, y_pred)))

Scikit documentation showing all possible metric modules

I don't think there are any iterations. Solving the linear regression problem is solving OLS which has a closed form and is calculated using SVD — Gilad Green, May 31 '20 at 09:11

score 0 · Accepted Answer · edited Jun 20 '20 at 09:12

As Pointed out by Gilad, no iterations are involved to calculate the linear regression problem using OLS . I am guessing here that you are trying to achieve the linear regression model using Gradient Descent. For the later one, number of iterations are defined prior the execution.

So to use the Gradient Descent, you can manually write the algorithm (Another link) or you can use the Ridge Regression or Lasso Regression from the Sklearn. Here there is an n_iter paramter by which you can achieve the number of iterations it took to determine the coefs.

class sklearn.linear_model.Ridge(alpha=1.0, *, fit_intercept=True, normalize=False, copy_X=True, max_iter=None, tol=0.001, solver='auto', random_state=None)

Attributes

n_iter : _None or ndarray of shape (n_targets,) Actual number of iterations for each target. Available only for sag and lsqr solvers. Other solvers will return None.

Note : You can define the solver as sag or saga to use the gradient Descent method.

Yes, apologies. Should have stated gradient descent - Thank you! — SheerKahn, May 31 '20 at 09:31

Scikit - Linear Regression - Extracting Metrics

1 Answers1