Group Lasso Math. Formulation Possible Typo

glm-tools / pyglmnet

Python implementation of elastic-net regularized generalized linear models

http://glm-tools.github.io/pyglmnet/

MIT License

279 stars 82 forks source link

Group Lasso Math. Formulation Possible Typo #397

Open cruyffturn opened 2 years ago

cruyffturn commented 2 years ago

Hello,

In the api, the regularization formulation is given as below:

\mathcal{P}_2 = \|\Gamma \beta \|_2^2 \ \mathcal{P}_1 = \sumg \|\beta{j,g}\|_2

The last summation is over g but not over j. Can it be the case that under-script "j" is a typo? I think the group lasso formulation should be the l2 norm of all coefficients in the group i.e. \|\beta_{g}\|_2?

Thanks!

jasmainak commented 2 years ago

I agree that the formulation is strange. The coefficients should be indexed over g where g \in G or an indicator variable. Can you propose a fix with a pull request? Thank you