Just to let everyone know, GSOC (Google Summer of Code) time is
nearing again. Last year we had two really good students and lots of
good proposals. Would love to see that continue. I'll post more info
when I have it, but here's some starter info:
For existing committers, If you are interested in mentoring, let me
For applicants, some things to keep in mind:
It's very important applicants demonstrate they are capable of working
and discussing ideas on the mahout-dev list during the application
phase. It simply is not enough to throw up a proposal on the GSOC
site, even a strong one, and expect to be selected. The Apache Way is
all about community. We want to hear the ideas and we want to discuss
them and we want you to be a part of the community. If you want
examples of that, see the archives from last year and our interactions
with our two students from 2008. Or, just look at any of the
interactions on the lists. Ask questions, help out, etc. If you
really want a leg up, demonstrate your proficiency, by creating a
small patch/demo that fixes/improves something in the current
implementations. See the How To Contribute section of the Wiki.
Lastly, before I get off my soap box, when applying, DO NOT claim to
be able to implement a whole slew of algorithms in one fell swoop. I
don't care how good you are (or think you are), it simply isn't
possible. Trust me. Even if you could (and you can't), the community
won't be able to keep up and then you won't be happy either. Instead,
pick one good idea and show a project timeline and a in-depth
knowledge of what you are proposing, including references, etc. If
you really think you could do more than one, instead propose items
that are "time permitting" and that build on what you have completed.
Demos and documentation are always good in this regard.
On Tuesday 17 February 2009, Grant Ingersoll wrote:
> For ideas on what we need, see: http://cwiki.apache.org/MAHOUT. To
> name a few: SVM, categorization algs, large scale graph ranking
> tools, maximum entropy implementation, collaborative filtering
> improvements (Sean?)
To name a few more: Algorithms for learning from sequential data (e.g.
identifying named entities in an incoming stream of text), algorithms for
learning rankings of items are also interesting.
If you plan to use Mahout as your platform of one of the various data mining-,
machine learning- or information retrieval challenges feel free to submit
your plan as GSoC proposal.
> For applicants, some things to keep in mind:
> It's very important applicants demonstrate they are capable of working
> and discussing ideas on the mahout-dev list during the application
A definite +1 from me. Discussing your idea before submitting the proposal
also helps you to get an idea of what exactly is needed, what is important to
keep in mind and to get a better proposal. So don't be afraid to post your
idea and refine it together with us.
> If you really think you could do more than one, instead propose items
> that are "time permitting" and that build on what you have completed.
> Demos and documentation are always good in this regard.
+1 It is not sufficient to submit a straight forward algorithm implementation.
Keep in mind that in order to remain maintainable you need to provide unit-
and integration tests for your work. In addition you need to provide examples
and demos so others can see how to use your work. Finally thorough
documentation of the algorithm itself, the implementation, its advantages and
limits is needed to evaluate it for commercial projects.
The abuse of greatness is when it disjoins remorse from power. -- William
Shakespeare, "Julius Caesar"
|\ _,,,---,,_ Web: <http://www.isabel-drost.de>
/,`.-'`' -. ;-;;,_
|,4- ) )-,_..;\ ( `'-'
'---''(_/--' `-'\_) (fL) IM: <xmpp://[hidden email]>