search

pbreheny / grpreg

Regularization paths for regression models with grouped covariates
http://pbreheny.github.io/grpreg/
34 stars 14 forks source link

`grpreg` objective function inconsistency #51

Open SzymonNowakowski opened 2 years ago

SzymonNowakowski commented 2 years ago

There is an inconsistency in documentation and vignettes for grpreg objective function:

  1. Instruction on page 15 defines the objective function as $$Q = \frac{1}{n} * loss + penalty$$
  2. Models vignette defines the objective function as $$Q = \frac{1}{n} * loss + penalty$$
  3. Penalties vignette, however, defines the objective function differently as $$Q = loss + penalty$$

I have tested the results of grpreg with glmnet on some examples and they are the same, so looking at glmnet objective function for Gaussian families I conclude that 1. and 2. are correct. Could you please update the Penalties vignette?