VisWeek 2012 started today, and it started with quite a bang. Here are some observations and highlights from the first day.
A great milestone is that as of a few hours ago, there were over 1,000 people registered. Surprisingly, over 700 people picked up their badges yesterday night and today, which probably makes this the busiest first VisWeek day ever.
Keynote Pat Hanrahan
Pat Hanrahan gave a great keynote at LDAV. He talked about a new approach to dealing with large data, which he calls divide and recombine. I’m not sure I understand it completely, but the idea seems to be to break data up into smaller chunks depending on different criteria, in order to essentially conduct experiments on the data. He argues that this is a more flexible way of dealing with large data, and a more efficient way of computing statistics than using pure Map/Reduce. Along the way, he showed some fascinating work on making R faster by implementing a just-in-time (JIT) compiler for it, and/or running R on top of Hadoop.
An interesting point Pat made in front of this audience that is traditionally very heavy on scientific computing was that high-performance computing (HPC) would soon be driven by business analytics rather than by scientific computing (which is what has been driving the field for many years). He also argued for looking over the fence and combining new ideas from cloud computing with HPC. This may seem obvious from an outsider’s perspective, but the two communities are currently completely disconnected, attend different conferences, etc.
There were some good discussions in the BELIV sessions today, on a variety of different topics. While they were all interesting, very little was really new. It’s not a new insight that we need to engage psychologists more, and there have been papers on evaluation methodology in the past. What is lacking is a bold next step that will move the field forward not just in terms of evaluation, but the way we think of visualization in the broader context of perceptual and cognitive research.
I only saw a small number of papers in BioVis today. What I found really interesting are the short primers at the beginning of every session. Since a large part of the audience have little or no background in microbiology, genomics, etc., these provide some background to understand what many of the papers are about. Even in other sessions, I can imagine a brief introduction being useful to provide some context, in particular ones on evaluation, etc.