Mainak Jas

@pavanramkumar this PR might fix our `learning_rate` woes. It's close to ready ...

Note also that in the community crime example, the R^2 for the grid search method matches what we find with reg path

@pavanramkumar with the fix in #249 the group lasso example gives reasonable R^2. So, it's fixing 2 or 3 examples. The only problem is that it slows down the convergence....

@pavanramkumar have you had time to look at this PR? It's related to our discussion today

I think line search does not make any assumption about convexity -- so it should apply to all the distributions. Sure I can add tests and report back some benchmarks

@pavanramkumar I rebased this PR with master. It does solve some issues. See above. But travis needs to be made happy ...

@pavanramkumar here is the [script to benchmark](https://gist.github.com/jasmainak/5ea97bde9f110e3071ac9e3ae9f0577b). And here is what I get on my computer for `alpha=1.0`: ```sh reg_lambda | Learning rate | glm.n_iter_ | time (s) 0.50000 |...

we're going to experiment with cython in the coming weekend hopefully. It's a 4 day weekend here ;) Neither me or @raghavrv have experience with `numba` though ... WIP PRs...

Thanks @pavanramkumar ! We _almost_ sat down with the intention of hacking cython today but ended up enjoying food in a delicious asian veg restaurant instead. I blame it on...

@pavanramkumar sorry I haven't been able to look into cython yet. But maybe it's a good thing that we don't rush premature optimization. Do we already have comparable performance (in...