Re: Mailet Development

[Robert Burrell Donkin <ro...@gmail.com>]

On Thu, Sep 29, 2011 at 4:27 PM, Adam Gibson <ae...@mtu.edu> wrote:
> Hi Robert,
>
> I have taken a look a closer look at the apache committers process and understand how it works now. I'll start working on some initial contributions and see what might be able to come out of this.

:-)

> Anyways,regarding The AI mailets as well as the problem space you mentioned with efficient ways of semi structured data are problems I'm interested in helping with. Initially, I will attempt to look more closely at what the other prject was you had mentioned was doing. I would want to start with few different machine learning algorithms including the standard (k-nearest-neighbors,SVMS). It also might be a bit excessive, but I would also like to experiment with an ensemble learning based approach to try to improve effiency.

+1

Ensembles have some very interesting qualities :-)

> Since the problem space is pretty easy in terms of having all the data, it might be worthwhile to look in to comparision metrics to know whether different algorithms are working well or not. I will do some more concrete research and come up with a more formal proposal here soon.

We're all volunteers so only be as formal as you feel comfortable
with. Here in James we use JIRA as an idea store [1], so that would be
one good way to start letting everyone know how you're thinking is
developing.

More later...

Robert

[1] https://issues.apache.org/jira/browse/JAMESMAILAI

---------------------------------------------------------------------
To unsubscribe, e-mail: server-dev-unsubscribe@james.apache.org
For additional commands, e-mail: server-dev-help@james.apache.org