Regularization

Simple Explanation of Regularization in Deep Learning and why it is Needed

L2 Ridge Regression

To be used when most variables contribute to the end result in some way or another.

L1 Lasso Regression

To be used when there are many unrelated/insignificant variables. L1 helps to completely get rid of these variables and retain only the significant ones.

Ridge Vs Lasso Regression, Visualized!!!

Very nice visualization explaining the difference between L1 and L2 regularization.