Two days at IBM Watson Research Center.

While the stock markets were collapsing, I was building relationship with eminent multimedia researchers at the Fourth Annual IBM Watson Workshop Emerging Leaders in Multimedia. Yes, somehow they felt that I am an “Emerging Leader in Multimedia”. One among the 12 students selected from world over (Although, the distribution is like 8 Chinese, 2 Indians, 1 Switz, 1 Greek) .

It has been an informative and entertaining experience. The first day was filled with student presentations, with me being the first one to present. Now, this might occur to you as a mere outcome of the pervasive randomness in the universe, but I have reasons to believe this was not the case. The problems I work on and the way I attack them is very similar to the approach adopted here at IBM. Researchers from the Multimedia Group, were present to attend my talk, and a couple of interesting points emerged out of the post-talk discussion. Primary of them being, how do we measure the confidence score of the classifiers (sorry for the jargon, but that would serve me as a reminder). There were some other interesting talks, but since this was a multimedia workshop, it brought people from sound (and video) analysis together. My ignorance in sound processing, took the worst of me, putting me into the “why am I here” mode. Anyways, post talk we went out for dinner. As usually with any multicultural dinner, the topics ranged from food to culture to weather to language, from India to America to China to Greece to Serbia. I enjoyed it. Finally the end of first day was in my istyle :). Our hotel was a good three miles away from the dinner spot, with most of the way back passing through isolated wooded areas, surrounding a beautiful lake. I decided to explore the village of Tarrytown (were we had our dinner), and walked back to our hotel. Saw two deer on the way. Tarrytown is also famous for being the place in “Legend of Sleepy Hollow”. Infact, the northern parts of this village renamed itself to “Sleepy Hollow” in 1996.

Today, the second day of the workshop, we were taken on the tour of the IBM’s Industry Solution Lab, were the newest IBM technologies were displayed for customers. However,  most of the technologies they showed did not seem to be the next generation technologies, probably because they were meant for customers and not for “Emerging Leaders in Multimedia”. In the afternoon we had a small group discussion on “Multimedia”, a part of which I missed, as I was supposed to meet one of the researchers one-on-one. I met Jelena Tesic, who walked me through her life at IBM, and the kind of vision related activities done here. Post lunch I met another two researchers, Nalini and John, who were working in biometrics. They had missed my presentation, so I gave them a quick tour of what I do. It was interesting to see John actually being able to understand every word that came out of my mouth. In short he understood what I was doing, which is rare, and his comments here and there gave me some ideas on how to improve upon the presentation. Again, for my benefit, I should have a slide that clearly shows that at the first level the semantic classes are learned based on their appearance model, where as at the second level, on their contextual attributes. He also told me that the notion of polysemy is also known as Perceptual Aliasing (a cool term to use in the papers).

In all, I think it was a fruitful endeavor.  I would be spending the rest of the weekend at Roti’s place with Bhatele, Ul and sardar giving company. Interesingly, Aman is also in New York, and this would be the second time I would be mixing school friends and IITK friends together. I just hate that :).