Glad to see you guys like it. :) You're right, at the moment you can't get predicted probabilities. It's something I'll be working on in the near future once I get a couple of other features out of the way (need feature selection + some variant of split-combine-apply transformer).
In terms of GLM.jl, I do intend to integrate it, just not sure when. Now that you mention it, I'll probably open up an issue soon for features that will be developed but are not major enough to be in the future work document. On Sunday, April 6, 2014 12:48:21 PM UTC+10, Samuel Jenkins wrote: > > Hi everyone, > > I've been having a go at Julia the past couple of weeks and am really > enjoying it! > Anyway, I'm developing a heterogeneous ensemble learning library called > Orchestra, of which you can find here (also in METADATA.jl as Orchestra): > https://github.com/svs14/Orchestra.jl > > It wraps around existing machine learning packages in Julia (at the moment > DecisionTree.jl + LIBSVM.jl), while also providing ensemble learners such > as stacking. I'd be happy to integrate more. > There's a long way to go, currently it has been tested for classification > tasks with numeric features and nominal labels. > I'd also like to (if it's not unbelievably difficult!) integrate Python's > scikit-learn and R's caret machine learners in the near future, that would > be extreme! > > Have a look if it interests you! > > Sam >
