Mainak Jas

did you look at the convergence plot? I think our solvers are not as efficient / optimized as celer or scikit-learn. You might be running out iterations.

@GJBoth is this still relevant for you? You might want to use the latest version to see if it fixes the problem

Can I ask if you are using the development version? And which solver are you using? It's a known issue for the cdfast solver. See: https://github.com/glm-tools/pyglmnet/issues/275 and https://github.com/glm-tools/pyglmnet/issues/289

hi @inferentialist sorry for the delay in getting back to you. I finally had a moment to look at this but I can't reproduce your problem. See my code here:...

If the R version of glmnet is based on it, I think it sounds like a great idea. What solver does it use? By the way if you are working...

hi @inferentialist I had a brief look at the paper you mention, it sounds promising. But we will need some work to integrate it into the current codebase. I'd be...

+1 although I would do it after `0.2` and before the paper. I think it will need some work ... and would be good to involve n > 2 people...

I wouldn't change the class signature too much any more. We made one big backwards incompatible change with this release (GLM and GLMCV). Now we should try to keep the...

I think the issue is that it's not complaining when there is a pull request. Try making a pull request where you make `flake8` errors in `tests/` and Travis does...

@pavanramkumar do you want to look into this? I guess we could start from [miniconda image on dockerhub](https://hub.docker.com/r/continuumio/miniconda/) and then copy the install commands from travis or circle