[mlpack-git] [mlpack/mlpack] Density Estimation Tree made sparse-enabled (#802)

[Ivan Georgiev]

Ok, I'll make a PR for sparse's sort as it is now - I'll try to add some tests shortly and you can choose when to get it.

Just to comment on sparse matrix loading - I'm sure it's easy to extend the Arma's `diskio` functions (`save_coord_ascii` can be used directly, but `load_coord_ascii` will need splitting - the size deduction part from actual reading). 

The more interesting question, I'd say, is how MLPack handles different input via `data::Load`, depending on the filename - pretty much the way it does not - but with the ability of initializing a sparse matrix, when that type of file is provided. This runtime flexibility is not very template-friendly, but my idea was to use variants. Something like `typedef boost::variant<arma::mat, arma::spmat> Dataset`, i.e. an encapsulation of different matrix types. There is a runtime mechanism to query what exactly type is the matrix read. I haven't implemented this, however - just sharing the ideas.

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/mlpack/mlpack/pull/802#issuecomment-257620102
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack-git/attachments/20161101/1bde1b0a/attachment-0001.html>