[Noisebridge-discuss] Machine Learning seminars?

Josh Myer josh at joshisanerd.com
Wed Dec 10 22:53:33 UTC 2008


On Wed, Dec 10, 2008 at 02:11:51PM -0800, Praveen Sinha wrote:
> I'm for sure down for it -- maybe we should get a night together in the
> upcoming weeks for all of us stat/machine learning nerds to sit down and
> implement something? 

It sounds like there's a lot more depth in this crowd than I was
anticipating, which is both awesome and a little intimidating (my
formal background is spotty, I just kind of use these things).

As for what to hack on: It might be good to have a quick BOF session
one night.  Collect notes of people's interests, and brainstorm some
project ideas.  After that, we could spend some time figuring out what
we could do.

Personally, I'm interested in helping remove some of the mystique
surrounding ML.  Feature selection is always going to be a bit of
voodoo, but I would love to build a cheatsheet for non-practitioners.
Basically, a decision tree, given the features of your data, here's a
set of models you should try in Weka.  Making weka appropriately
user-friendly in displaying the output of this is a separate task, and
likely full of visualization goodness.

> We need some cool datasets to work off of!  Anyone have any ideas?

The first resource to plug is (theinfo).  Aaron Swartz started it up
some time ago, and it seems to be going strong enough, but could
always use more contributions.

http://theinfo.org/

Otherwise, if we find any new datasets, we should just add them to
theinfo!
-- 
Josh Myer   650.248.3796
  josh at joshisanerd.com



More information about the Noisebridge-discuss mailing list