I am looking into running regression on a multivariate data set. I am looking into different ways to scale my data: standardization, L2 and L1 normalizations.

In what case would you use which method?


  • $\begingroup$ By L2 and L1 normalization, do you mean regularization? As in shrinkage with ridge and LASSO? $\endgroup$ – Frans Rodenburg Nov 8 '17 at 4:19
  • $\begingroup$ L2 and L1 are used for regularization but can also be used for scaling your dataset. en.wikipedia.org/wiki/Feature_scaling $\endgroup$ – ninjaSurfer Nov 8 '17 at 4:27
  • $\begingroup$ ...by which you mean, for L2 that you scale all your feature vectors to lie on the unit sphere? And by L1 on the unit L1-norm-surface? $\endgroup$ – bibliolytic Nov 9 '17 at 6:37
  • $\begingroup$ exactly. it is used many time when preprocessing data for machine learning. I couln't however find an analysis on when each method works best. $\endgroup$ – ninjaSurfer Nov 10 '17 at 4:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.