Adding Group Lasso Penalization

Hi,

First of all, thank you for the great work on the GLUM package — it's nearly perfect for my needs!

I’m currently working on my master's thesis and was wondering if Group Lasso penalization is something planned for future development (related to #478)? This feature would be very helpful for my use case, and I’d love to know if it’s on your roadmap.

I’m planning to use GLUM for two main tasks:

1. Standard GLM Regression
   My input dataframe X contains variables that are naturally grouped — for example, categorical variables. I’d like the regularization to treat these variables as groups, penalizing or selecting them as a whole, instead of selecting/deselecting individual categories within a single variable.

2. Generalized Additive Model (GAM)-like Regression
   I’m introducing spline transformations to my numerical variables to capture non-linear relationships. Each variable is represented by a set of spline basis functions. Ideally, the penalty would apply to all splines for a variable as a group, to either select or drop the entire set together.

Note that P2 matrix allows me to smooth my splines, then no problem on this side as shown by this notebook.

Given this, I have two questions:

• Is there a way to achieve this behavior with the current version of GLUM?

• Or, would it be better for me to fork the code and adapt it to my specific needs?

Thanks again for this great tool — it's been very helpful in my research!

Best

Start by reviewing related issue #478 and the current GLUM penalization capabilities. Use the linked penalized-splines notebook to understand the stated spline use case; the work is complete only when grouped penalties can select or drop categorical groups and spline bases together.