Roman Cheplyaka

Yes, I also noticed that. But I don't think we should regard that as a standard; see e.g. [this stackoverflow discussion](http://stackoverflow.com/q/6837628/110081). Another question is whether `unlines` should be changed, too......

Shelling out would be fine with me; e.g. in R's data.table you can say `fread("xzcat file.csv.xz")` and it'll run that command and read its output as a csv file. An...

Thanks for the response Bob, that's very useful to know. However, the two programs do give the same (incorrect) answer for optimization. Also, I just checked that R's optim finds...

Unfortunately, I can't seem to reproduce your results. If I don't supply `init` as in your comment https://github.com/stan-dev/stan/issues/2511#issuecomment-378731571, then regardless of the tolerances it arrives at the right answer about...

@mikeizbicki I'm also running into this. Could you fix it?

Well, yes, but that set subhask to an old commit well behind master. Was that your intention? Also, I just noticed that HLearn's stack.yaml refers to subhask on hackage, not...

> There is an old version of subhask on hackage, but the version that HLearn should be using is whatever commit the submodule points to. Gotcha. This is not the...

At the moment I'm dealing with multi-class classification. I vaguely know that I can use logistic regression for predicting each class separately, but I don't know how well it works...

I pushed a second, independent commit (let me know if you'd like me to split this into separate PRs). Hopefully the haddock there is self-explanatory. I've always assumed that freshen/freshenAll...