[mlpack] GSoC 2014 Simulated Annealing Project

Ryan Curtin gth671b at mail.gatech.edu
Tue Feb 25 12:34:44 EST 2014


On Tue, Feb 25, 2014 at 11:31:06AM +0530, Abhishek Laddha wrote:
> Hi all,
> I am Abhishek Laddha a 3rd year undergraduate student majoring in
> Mathematics and Computing from Indian Institute of Technology Delhi (IITD),
> India. I am really interested in field on machine learning and regularly
> taking the similar courses in my undergraduate studies. I am familiar with
> C++ and some of machine learning techniques.
> 
> I would like to work on project "Simulated Annealing Optimizer" for gsoc
> 2014. I have already used simulated annealing technique in my project.
> Simulated annealing is a search in the solution space of constrained
> optimization problem which is good in avoiding local minima but it is slow.
> I have found the the some paper of using simulated annealing technique in
> K-means algorithm and linear regression. Could you more elaborate on this
> project ?

Hi Abhishek,

Can you link me to the paper where simulated annealing is used for
k-means and linear regression?

To understand the project better, take a look at the optimizers already
present in src/mlpack/core/optimizers/ to get an idea of how they are
written.  Most importantly, they are generic -- so, they can work with
any potential function to be optimized, assuming that function
implements the Evaluate() function and Gradient() function (although for
simulated annealing, Gradient() is not necessary).

> I have downloaded, build and  gone through the some mlpack tutorials. Could
> you give me direction or provide resources for how to start working on this
> project?

If you are interested in contributing, a list of bugs can be found here:

http://www.mlpack.org/trac/report/10

To get an idea of how mlpack is actually used, it would probably be a
good idea to download some datasets and perform tasks with them using
mlpack; for instance, you could find a dataset to perform regression on
with both linear_regression and LARS.  You could also perform
nearest-neighbor search, furthest-neighbor search, max-kernel search,
and other tasks like that.

Let me know if I can answer any more questions.

Thanks,

Ryan

-- 
Ryan Curtin    | "Weeee!"
ryan at ratml.org |   - Bobby


More information about the mlpack mailing list