The VisWeek keynote is just over; the fast-forward and the first VAST session are now underway. There’s more VAST in the afternoon, and also what I think will be a great panel.

I thought that the keynote was somewhat disappointing. It seems wrong to snark at an invited speaker, but I wish his talk had a little more justification for the visualizations he presented; the animation of the neural network process, for example, seemed to me to use visualizations for precisely the wrong reasons: it showed something, it looked good, but it never became clear what the visualization taught the users. What’s the point, then? Am I missing something here?

One paper I’m looking forward to seeing this morning is Albuquerque et al’s Perception-Based Visual Quality Metrics. The interesting bit here is that they use perceptual experiments to derive a metric through explicit energy minimization. This is neat: the paper uses human subjects where computers can’t do a very good job, and computers where humans do a good job. This is something that visualization papers tend to get wrong: because the community is good at creating depictions of data which are intelligible and attractive, we tend to forget that computers are actually great at computing things!

In the afternoon, there’s Orion, Heer and Perer’s system for manipulating network data. The only systemsy tool I know for generating and processing network data is GVPR, and it feels, ahem, slightly outdated. It certainly gets simple jobs done, but AWK expertise seems to be falling out of favor, and Heer has a good feel for computational building blocks. I’m looking forward to that talk.

In the afternoon as well, there’s van Wijk, Ware, Demiralp and Laidlaw’s aptly named panel: “Theories of Visualization: are there any?”. This continues the long-standing debate on how to appropriately define and measure the goodness and utility of a visualization. Old readers will remember my discussion on the topic at the time of Laidlaw’s capstone some years ago. This problem continues to be important, continues to be hard, and continues to be completely open.