I was wondering about the assumption for the distribution of the features for regression models like Lasso regression or ridge regression. why it's better to have features with Gaussian distributions ?