[mlpack] GSoC 2014 simulated annealing optimizer

[Zhihao Lou]

Hi Ryan,

Thanks for the help. I'm starting working on the SA optimizer now.

What's the project's preferred random number generator for uniform [0,1]
scalar? Or should I use arma::randu<vec>(1) ? Or my favorite random
generator?

Zhihao Lou


On Mon, Mar 24, 2014 at 3:05 PM, Ryan Curtin <gth671b at mail.gatech.edu>wrote:

> On Sun, Mar 23, 2014 at 09:18:39PM -0500, Zhihao Lou wrote:
> > Hi Ryan,
> >
> > Thanks for the reply. It's my own fault missing the deadline. I should
> have
> > known better.
> >
> > On the other side, would you please give me some instructions on how
> should
> > I start to contribute? Should I started with some branch or the trunk? My
> > guess is my code will belong to src/mlpack/core/optimizers/sa, and the
> > interface should closely resembles those existing optimizers. But what's
> > the proper way to get information about the problem being optimized, like
> > the number of dimension, or the range of each parameters? I can always
> get
> > these as parameters of the constructor, but it is better to get these
> > directly from the function itself, if that's possible.
>
> Hi Zhihao,
>
> You can take a look at the other optimizers present in
> src/mlpack/core/optimizers/ (in trunk) to get an idea of how these
> questions are currently handled, and to get an idea of the interface the
> simulated annealing implementation should implement.  The other
> optimizers generally get the size of the problem from the
> GetInitialPoint() method.
>
> Thanks,
>
> Ryan
>
> --
> Ryan Curtin    | "Exterminate all rational thought."
> ryan at ratml.org |   - Bill Lee
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cc.gatech.edu/pipermail/mlpack/attachments/20140324/031e1518/attachment-0003.html>