MLSG: Fourth meet

[ambarisha b]

Hi,

This week we will be discussing support vector clustering. Ajithesh will lead the discussion once again. If anybody can talk about any topic, please volunteer. Also post the reference links on this thread. Here are a couple of ideas from our discussion during the last meet

Limitations imposed on Kernel functions
R, how to link R libraries and use them
Naive bayes
Any research paper extending our previous discussions

This week lets start the idea-implementation thing. So, brainstorm for ideas. We can discuss the idea during the meet and a group can start work on its implementation. Infact, posting an abstract sorta thing for the idea, on this thread would be even better.
Jump in with any other inputs as well.

//Ambarish

[Ajitesh Srivastava]

Few links related to Support Vector Clustering:

http://jmlr.csail.mit.edu/papers/volume2/horn01a/rev1/horn01ar1.pdf
http://www.springerlink.com/content/xx645pm2n21m6268/fulltext.html

I feel we should not go deep into support vector clustering but should discuss the basic ideas and then perhaps move to Multiple Kernel Learning and other things.

[ambarisha b]

A small change in today's meet timings.  Its from 10pm - 11pm in 1232. Hope that ain't much trouble. Please try to make sure everyone gets the message.

Cheers
Ambarish

[Shaleen Sharma]

The meet today stands postponed till further notice. Apologies for the late info.

--
Regards

Shaleen Kumar Sharma

Undergraduate, BITS, Pilani

[ambarisha b]

There was some electricity problem in room 1232 today and 1233 was occupied for mime practice, so we couldn't have the session. It has been rescheduled to sunday 8PM. If somebody isn't ok with the timings, mail me directly and we'll sort it out.

Regards
Ambarish

[ambarisha b]

Change of plans. The room could not be booked today. As tommorrow we have ASIG, we have to move to Tuesday. That will disturb the schedule. Instead of that, we shall have a 2-hour long session on Friday itself. Sorry for the inconvenience.

Ambarish

[Rishabh Mehrotra]

For a 2 hour session Support Vector Clustering doesn't seem enough. Regarding Multiple Kernel Learning, there a bit of Optimization(of course, non-BITSian Opti) involved. Any suggestions for another topic for a 2 hour session?

--
Rishabh.

[Sahil Shah]

We could discuss Hidden Markov Models and their applications in time series analysis like gesture/speech recognition.

--
Sahil Shah

[ambarisha b]

This is a remainder of today's meet @ 8PM in 1232