Solving a constraint lasso

Specific Domains Optimization (Mathematical)
1

Hi,

Recently i had the problem of solving a constraint lasso, much like the one described in this paper. The paper refers to a julia package, ConstrainedLasso.jl, that is unfornutely not updated since julia 0.6. With some help from the forum I gave a shot at updating it but without sucess so far.

Then I had a thought: Is there in the current Julia packages some infrastructure that could solve this kind of lasso problem ?

I found GLMNet.jl, but which does not allow linear constraints, same thing for Lasso.jl and LARS.jl, sadly.

2

You could probably extend this example:

https://jump.dev/Convex.jl/stable/examples/general_examples/lasso_regression/

Might not be the quickest, but works.

3

Thanks a lot. After a quick learning of Convex.Jl (Awesome stuff Btw), i managed to add my constrained and run a constraint lasso as I wanted.

Indeed, this is quite slow to obtain the full path as I must solve it again for each regularisation parameter, and i suspect that the solver is not optimized for this kind of stuff.