[mlpack] GSoC 2016 - Project Questions

[Ryan Curtin]

On Tue, Mar 08, 2016 at 06:43:13PM +0000, Yannis Mentekidis wrote:
> Hello,
> 
> Thank you for the clarifications, I managed to run specific tests after
> that.
> 
> I'll look at Datar's paper and your comment, and hopefully I'll be able to
> have a draft idea about a testing procedure soon. Should I get back at you
> with it to discuss it further, or should I just start coding away and see
> how it goes?

If you're confident that the test you're making is a reasonable test of
correctness, then it's better than what's there now.  So please, go
ahead and give it a shot! :)

We can work out minor details when you submit a PR (like style issues or
whatever).

> I believe the dataset used for this shouldn't matter that much, so we
> probably will not need to include a different one. I'll probably toy around
> with the iris data for now.

I agree; other options might include the slightly larger 1000-point
random 3-dimensional dataset (test_data_3_1000.csv) and the vc2
(vertebral column) dataset, in vc2.csv.  There are a couple more that
might be useful, but in reality it's not so big of a deal.  In fact it
wouldn't be hard to run whatever test you come up with for a couple
datasets.

> Thank you very much for the input!

Sure!  I'm glad that I can be helpful here. :)

Thanks,

Ryan

-- 
Ryan Curtin    | "I am a golden god!"
ryan at ratml.org |   - Russell