On Mar 19, 2008, at 6:18 AM, Khalil Honsali wrote:
Hello all,
I am writing you to inquire about the projects posted for Google
Summer of
Code (GSoC).
Please consider this as step[0] of the process. If this mailist is
not the
right context for this topic please disgard this message and kindly
suggest
appropriate context.
First, let me introduce myself, I am master course student at
Nitech.ac.jp;
I just finished my first year dedicated to courses and about to
start the
research project. I have been around hadoop since 2007.end; though I
define
myself as beginner user, to be honest, I have a basic understanding of
Hadoop architecture, how to use it, its API and map/reduce programs.
My actual research topic/interest is targetting efficient index
distribution
for a digital library system; in which I wish to let hadoop digest
citations
and generate indexes/metadata to be distributed around HDFS. Hence,
I see in
GSoC as an exellent opportunity to practice with Hadoop, a good
excuse to
benefit from experienced mentors and why not eventually output a
publication?
I understand that in case my participation is welcome, first I have to
discuss then submit to ASF::Hadoop a formal idea proposal before
applying to
Google. Through the ASF GSoC page, one interestingly challenging
project is
about Hadoop:Mahout's machine learning algorithms, I am more
specifically
attracted to Neural Networks problem, as I enjoyed the one and only
course I
had on the matter (we just had to make a simple BBP on a input sine
wave
approximation); though I believe such a problem might fit more
naturally
into Dryad's acyclic graph model than mapreduce....
My questions are:
= what are qualifications to join ASF's Hadoop GSoC ?
= in that case are there specific guidelines to follow?
= more importantly what are your suggestions concerning the above
matters?
= and finally what's next from here?
I think most of the info is at http://wiki.apache.org/general/SummerOfCode2008
and the timeline at: http://code.google.com/opensource/gsoc/2008/faqs.html#0.1_timeline
FWIW, I think implementing a M/R Neural Nets would be great.
Essentially, what you need to do is make a proposal. I think you can
look at prior year proposals to see how those look. One thing to
think about is how much time it will take to implement, so you may
want to put together a progression of tasks, w/ some maybe even being
optional depending on how the earlier ones go.
Cheers,
Grant
