[mlpack] mlpack Digest, Vol 13, Issue 27
Andrew Wells
andrewmw94 at gmail.com
Fri Mar 14 12:53:27 EDT 2014
Hi Ryan. I'm happy to talk about my previous experience, I just wasn't
sure whether you wanted to read about in my application or on the
mailing list. Here goes:
Robocode is an AI programming competition, where I have one of the top
competitors (http://robowiki.net/wiki/Gilgalad). The robot uses the kNN
algorithm for data analysis, which is done under strict time
limitations. To deal with this, I implemented a tree that is similar
kD-tree, but with minimum bounding rectangles. (source code is in the
Mallorn package here:
https://sites.google.com/site/awusa94/robocode/aw.Gilgalad_1.99.7.jar).
Because the trees are built dynamically, I unfortunately don't have
experience with bulk-loading. However, the tree is highly optimized,
based on my own experiments and discussions from the robocode community.
(I can provide a plethora of examples of these discussions if you want.
For starters I have:
http://robowiki.net/wiki/User_talk:Rednaxela/kD-Tree ,
http://robowiki.net/wiki/User:Duyn/kd-tree_Tutorial , and
http://robowiki.net/wiki/User_talk:Skilgannon/KDTree). Of course, some
of these optimizations are Java specific, but many are not. For example,
traversing the tree without recursion gives a small speed-up by
eliminating the stack overhead, and some of our experiments indicate
that descending the tree with a "depth first" strategy and then checking
the other branches in a "breadth first" order can give better results
than the standard descent ascent approaches.
I hope that gives you a good idea of my experience with robocode. I go
into more detail in my application, but if you want more here, feel free
to ask.
On 03/14/2014 12:00 PM, mlpack-request at cc.gatech.edu wrote:
> Send mlpack mailing list submissions to
> mlpack at cc.gatech.edu
>
> To subscribe or unsubscribe via the World Wide Web, visit
> https://mailman.cc.gatech.edu/mailman/listinfo/mlpack
> or, via email, send a message with subject or body 'help' to
> mlpack-request at cc.gatech.edu
>
> You can reach the person managing the list at
> mlpack-owner at cc.gatech.edu
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of mlpack digest..."
>
>
> Today's Topics:
>
> 1. GSoC project idea (Andrew Wells)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Fri, 14 Mar 2014 12:00:13 -0400
> From: Andrew Wells <andrewmw94 at gmail.com>
> To: mlpack at cc.gatech.edu
> Subject: [mlpack] GSoC project idea
> Message-ID: <5323278D.7080609 at gmail.com>
> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>
> Hi all, I'm Andrew Wells, I am in the process of applying to work
> on mlpack for Google Summer of Code, and I wanted to introduce myself.
> I'm hoping to work on optimization of the tree traversal code in mlpack,
> as this seems to be the area where I have the most previous experience.
> I was wondering if you could give me some more information about this
> project. Also, I think it would be helpful for me to first submit
> another tree, say an R*-tree, to the mlpack code base to familiarize
> myself with it. Am I free to combine project proposals like that?
> Thanks for your time and I hope to work with you over the summer.
>
>
> ------------------------------
>
> _______________________________________________
> mlpack mailing list
> mlpack at cc.gatech.edu
> https://mailman.cc.gatech.edu/mailman/listinfo/mlpack
>
>
> End of mlpack Digest, Vol 13, Issue 27
> **************************************
More information about the mlpack
mailing list