Thursday, May 03, 2007

ICDE 2007 Keynotes

IEEE 23rd International Conference on Data Engineering
17-19 April 2007 - The Marmara Hotel, Istanbul, Turkey

Keynote: Yannis Ioannidis

On the first day, Yannis Ioannidis gave a presentation on Emerging Open Agoras of Data and Information. He spoke about competitive markets, sharing of information and the motivation behind allowing the dissemination of information between parties through bargaining or other types of negotiation. I think the use of the term "Agora" was particularly appropriate based on our location (Istanbul). At the end of this presentation, he was asked to compare an Angora model with that of a Mall (US shopping center). Fun times.

Keynote: Ricardo Baeza-Yates

On the second day, Ricardo Baeza-Yates presented Challenges in Distributed Web Retrieval and discussed the issues of increasing Web data and using distributed machines to cope with scalability.

Keynote: Laura M. Haas

On the final day, Laura M. Haas spoke about Information for People, a presentation targeted at ICDE researchers to consider the HCI component of research development. I think this was the presentation I remember most vividly. She gave visual presentations and examples of some algorithm researchers who used GUI components to their research for representation of results. This visually helped anyone who was unfamiliar with the data to make clearer judgements. For example, instead of just looking at the numbers, std, and means, plot them and allow for dynamic modification of graph to anaylse trends in numerical data better. This of course was done in colour. Pretty. And she was right, it did help the comprehension of what was going on a lot.

However, I'm not sure how well it was received on my side of the room. There was some general consensus that while it's a nice idea to make our algorithms "pretty", the reason why HCI and Data Management were two distinctive areas was because the are completely two distinctive areas. I don't think there's another way to put it. So I think the best example I have is the OSI model. You're trying to merge the physical layer with the application layer. It's not quite that extreme but it's close. At ICDE, there were sessions and industry reports on more efficient caching techniques. How do we make that more GUI/user friendly? I think at this point in time in research, there is still a strong distinction between these things and there needs to be, otherwise the scope you start dealing with is just enormous. The point is abstraction and to work with manageable portions. It's a nice idea, but I'm not sure how practical it would be.

No comments: