Day 9: Edinburgh, Scotland (IJCAI Conference)
09:00 - 17:30: The first day of technical presentations at the 19th Biennial International Joint Conference on Artificial Intelligence (IJCAI-2005).
This morning's invited talk was by Alison Gopnik of Berkeley on some work she's doing with my collaborator, Clark Glymour, at CMU. Specifically, she talked about how children's exploratory experimentation is really a kind of causal influence rather than a defect and how we should think of kids as little causal inference machines. I really have to get her videos - besides being scientifically interesting, they are adorable.
The title was "Children, Causality, and Bayes Nets" (Gopnik is a child psychologist) and it was a very interesting talk. Afterwards, I asked her about active learning and whether "learning with a helpful teacher" (Salzberg, Kasif, Goodrich and Delcher, 1991) could be seen as a type of normative theory for causal interventions in educational psychology and the theory of pedagogy. Does anyone know if we have child psych and learning theory people in k_state's psych department?
I met Alison Gopnik and George Luger (author of a well-known textbook whose first three editions were co-authored with Stubblefield, and which has been supplanted by Russell and Norvig since around the second edition). During the day, I said hello to Lise Getoor, Avi Pfeffer, Dan Roth, Tom Mitchell, Ray Mooney, and Pedro Domingos, and talked with Jin Tian, David Page, and David Jensen.
In the relational learning track, Taisuke Sato of Tokyo University gave a very interesting talk on PRISM, a system that integrates Prolog with statistical relational learning. Next was a paper by Page and Ramakrishnan (both at Wisconsin) with some students and medical informatics collaborators, on using relational learning to improve on diagnostic MRI (mammogram readings by radiologists).
Before going to another track, I shook hands with Sato and said it was good to see him again. I wonder if he remembers that he's the person who coined the term "CRUSH HUGIN".
Next up was Carlos Guestrin's track on adaptive planning. It turned out that Peter Stone had swapped with him so he could be down at the Trading Agent Competition on the exhibition floor later. We talked briefly about work of Steve Gustafson, one of my graduate alumni, and Peter's paper with Risto Miikkulainen.
In the afternoon, I talked with Ric Crabbe of the U.S. Naval Academy, who chaired a robotics event at IJCAI-2003 in Acapulco and is also a recently-tenured faculty member. Let me tell you, I don't envy anyone who runs those robot events, much less in Mexico, where it's hard to take robots! Dave Gustafson elected not to for that very reason.
Just as Dave Page and I sat down to catch up, a bald dude stepped in front of my and introduced himself. It was my old friend and classmate, Dan Gaines (contributor or inspirer of many items from "101 New Things Not to Do at Your Ph.D. Defense").
The second invited talk was by Andrew Blake of Microsoft Research Cambridge, on "Visual Tracking of Objects in Motion". He surveyed about twenty-five years of research in about an hour and went over a lot of basic techniques. It was a good tutorial; I just wish there had been a little more on current work.
In the afternoon, I talked with Jesse Davis (one of Page's students), went to the exhibition and picked up brochures, cards, pens and widgets from companies (BT), publishers (IOS Press, Taylor and Francis, Elsevier, Cambridge University Press, Oxford University Press), and professional societies (BCS), and conferences (IJCAI-2007 in Hyderabad, WWW2006 in Edinburgh). Then I went and had a short talk with Randy Goebel and a long talk with David Jensen and skipped the first tracks on Information Extraction, Learning and Knowledge, Robot Perception and Learning, and Bayesian and Theoretical Learning.
17:30 - 19:00: Go home and sit for a bit.
19:00 - 20:30: Dinner at The Office. British food at last! £18.32 (£20.50 tip included).
- me: haddock and chips, half pint of Deuchars IPA
- Banadad: bangers and mash, half pint of 80 Shilling Scottish Ale
- Banamum: BBQ wings with sweet chili
Now, this all sounds very orderly, but if you've ever seen Chinese families eat, nobody's food stays on his or her plate alone. Banamum and I ended up having half of the bangers and mash and the Banafolks shared nearly half of the fish and chips. It's the whole concept of ordering dishes rather than entrées.
The Office, viewed from across the street, and an advert for the ale I had at dinner.
20:30 - 00:10: Home again for some unwinding.
Banamum checking investment news on Hirilonde, my new Dell Inspiron 6000.
- BBC2: Vacation Disasters - ha!
- Process some pictures from yesterday.
- Send and answer mail.
- A 45-minute discussion about London. We decide not to try to go see the sights, but will proceed directly from Gatwick to Heathrow. Bah.
00:10: Lights out.