18:00 < islamfaisal> Hi all, I would like to participate in Google Summer of Code this year with mlpack, is it too early to discuss project ideas or may I propose a few ideas now?

18:01 < naywhayare> islamfaisal: it's not too early to discuss ideas :)

18:01 < naywhayare> what do you have in mind?

18:02 < islamfaisal> Thanks for fast reply. I skimmed mlpack documentation but did not find algorithms for feature selection already implemented, am I correct?

18:07 < naywhayare> hm, maybe you could use some algorithms we have for feature selection... or I guess that's more feature extraction

18:07 < naywhayare> one could use PCA or kernel PCA for feature extraction

18:07 < naywhayare> but you are right, we don't have any explicit feature selection algorithms

18:10 < islamfaisal> What do you think can be added as a start, corellation based feature selection?

18:11 < naywhayare> I'm not actually an expert on feature selection

18:11 < naywhayare> but maybe some set of feature selection algorithms could be added to mlpack

18:12 < naywhayare> I think for a compelling project, the amount of work would have to be fairly significant, so it may be worth considering multiple algorithms

18:12 < naywhayare> in the past, GSoC projects that have implemented new algorithms for mlpack also include making sure they work with other algorithms, making sure they are as fast as possible, and writing detailed documentation and tutorials

18:15 < islamfaisal> Thanks for yor time. You are right, implementing a group of feature selection algorithms maybe a good candidate for GSoc projects. It was nice to meet you :)

18:16 < naywhayare> yeah, nice to meet you too :)

18:16 < naywhayare> I am usually in here, and I'd be happy to talk more about project ideas as time goes on

18:17 < naywhayare> but keep in mind that mlpack hasn't even submitted an application to be a mentoring organization yet :)

18:17 < islamfaisal> I am sure you will submit :)

18:17 < naywhayare> so it is not guaranteed that we will be a mentoring organization

18:17 < naywhayare> ah, yes, definitely, we will, but it's up to Google if they will accept us

18:17 < islamfaisal> aah I got that

18:17 < naywhayare> although, Google is interested in machine learning, so I think that they will continue to be interested and let us in for a third year...

18:18 < islamfaisal> meanwhile, until proposals are announced

18:18 < islamfaisal> How should I start

18:18 < islamfaisal> prepare a documentation, implement a basic algorithm or do what?

18:18 < naywhayare> hm, so the first thing to do will be to get familiar with the library by downloading it, building it, and maybe using it for some simple machine learning tasks

18:18 < naywhayare> and get an idea for what it's like to use mlpack, and what the code looks like

18:19 < naywhayare> there are a number of tutorials you could go through on the website

18:19 < naywhayare> once you're done with that, you could start contributing

18:19 < naywhayare> you can find a list of bugs on Trac (http://www.mlpack.org/trac/) or you could implement a new algorithm like you suggested

18:20 < naywhayare> some of those bugs are very in-depth so they may not be good choices, but if you see a bug where you think "ooh, cool, I can do that", then I (or someone else) can help you through the fixing process

18:21 < islamfaisal> Great, I will start checking Trac and let you know if there is something I can help with.

18:21 < islamfaisal> Again nice to meet you and join mlpack :)

18:21 < naywhayare> :)

18:21 < naywhayare> good to meet you too :)