Elastic-Net Regression

Used when there are a lot of features where we dont know whether they are useful or not, so we can’t decide between Lasso Regression and Ridge Regression. That’s why one should use Elastic-Net Regression in this case.

It combines the strengths of both approaches by including a Regularisation term for both Lasso Regression and Ridge Regression.

So we have ${\lambda }_1\mid s\mid +{\lambda }_2{s}^2$ We can use simple K-Fold Cross Validation to calculate the best lambdas.

Resources

• https://www.youtube.com/watch?v=1dKRdX9bfIo