Elastic Net Regression

#machine-learning #interview

Though the name suggests Regression, it's a regularization method for preventing Overfitting

Elastic Net Regression

$\begin{aligned} L o s s & = l o s s_{p r e v} + L_{1} + L_{2} \\ = l o s s_{p r e v} + λ | s l o p e | + λ s l o p e^{2} \end{aligned}$

It is a combination of both L1 or Lasso Regression and L2 or Ridge Regression
- it will do both Feature Selection and decrease Overfitting

Related Notes