[mlpack] Regarding Google Summer of Code 2013

Ryan Curtin gth671b at mail.gatech.edu
Tue Apr 9 12:14:36 EDT 2013


On Tue, Apr 09, 2013 at 09:27:42PM +0530, Pramanshu Rajput wrote:
> I am an undergraduate student at Indian Institute of Technology Roorkee,
> India
> 
> I want to know about the possibility of project on adding new features to
> mlpack including Restricted Boltzmann Machines and some Genetic Algorithms
> ranging from very basic like Simulated Annealing, (Mu,Lambda) Evolution
> Strategy to some of the more advanced algorithms.

Genetic algorithms are entertainingly slow, so I am not sure how useful
they are in the context of mlpack, which aims to be a fast library.
Also, simulated annealing and genetic algorithms are not the same thing,
but they are both global optimization strategies (in some sense).

You are not restricted to only the ideas in the Ideas List (as it says
on that page), but if you are considering doing some other project, you
will have to detail it in your proposal.  Your proposal will have to
convince us that your idea is a fast algorithm that can be implemented
without additional library dependencies, that you know the algorithm
inside-out and can implement it efficiently, other implementations that
you seek to outperform, and then how it will fit into the existing
mlpack codebase (both from the perspective of code cleanliness and also
the bigger picture of how all mlpack algorithms fit together).

Hopefully this is helpful.  I was not entirely sure how to answer your
question because it was a little vague.

-- 
Ryan Curtin       | "The mighty hand of vengeance, sent down to strike
ryan at igglybob.com | the unroadworthy!"  - The Nightrider


More information about the mlpack mailing list