![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
jrtomIn which I temporarily make this sound more like a journal, or a research blog, than like the "hey, look at this cool and/or wrong thing!" postings that generally characterize this space.
Most readers of this here blog are probably aware that Megan and I recently took the hit show "Corwin!" on its first ever road trip, with several dates in and around Portland and Seattle. We anticipate lots of cute pictures with Faces Not Previously Appearing In This Film.
Before I left for KDD, I had a chance to catch up with my friend Danyel, and to attend our friend Mara's wedding reception (the timing of the latter having determined the timing of our trip). At the reception, there was a girl, perhaps 11 years old (the daughter of one of Mara's friends, whom I'd first met before she gave birth--oh, how the years fly) who attempted to recapitulate some of the basic techniques of social network studies in about an hour, with some amused guidance from Danyel and I. :)
Some of you may not have known that my mysterious absence of a few days in the middle of the trip was not due to the usual alternate-identity crime(fighting) spree, but to my attendance at a scientific conference in Chicago, called KDD (Knowledge Discovery and Data Mining).
The primary purpose of my attending this conference was to present a paper at one of the workshops which preceded it. (For all 2/3 of you who care, it was "EventRank: A Framework for Ranking Time-Varying Networks".) This is my first publication in a peer-reviewed workshop. (I have several prior publications, but all of them are either not peer-reviewed, or only "published" in the loose sense that my MSc thesis is available in the UBC library--I never did manage to get my MSc advisor to help me get (part of) it actually published. Maybe someday I will--it's a pretty good guess that no one will have bothered to scoop me in the meantime. :P :) )
The presentation of the paper went very well. I didn't have time to give a practice talk, or even to run through it ahead of time, but my advisor and I were told by a number of people that it was the best of the workshop talks, and several people spoke to me over the next few days about the possible applications of our framework to their problems. (I was amused by the fact that my advisor apparently became more impressed with the talk, and the paper, every time he thought about them thereafter.) Having attended a number of the other talks, we think that if we had done the work a few months earlier (i.e., prior to the conference submission deadline), it would have been accepted to the main conference exactly "as is"--ah, well. (Conference publications are, in my field, considered more prestigious than workshop publications.) Anyway, a somewhat expanded version will get published in a journal or another conference soon, I expect.
This will not be a big surprise to anyone, but authors of papers don't necessarily do the basic background research they need to do. (I'm probably guilty of this myself, but I like to think that I'm more forthright about acknowledging my own failures in that regard when they're pointed out to me.) At least one presentation was given by a researcher who seemed to be treasuring his ignorance, however. *sigh* If this happens to me, I hope that I will at least have the grace to admit that there might be something that I missed.
Of course, an important secondary purpose for attending a scientific conference--especially for one such as myself who hopes to be Making a Career Transition in the foreseeable future--is schmoozing. (My friend Danyel has stated in the past that schmoozing is actually the most important reason. I don't entirely agree with him, but he's got a point.) The well-schmoozed included people from various different academic institutions (Stanford, UMass-Amherst, U of Washington, Carnegie Mellon U), several people from Yahoo! including the new head of Yahoo! Research, some folks at Amazon, and someone from Google (who apparently can't hold a conversation with a possible prospective employee without asking an interview question--fortunately I answered it correctly).
I talked to a number of folks about JUNG. People continue to be impressed. It seems increasingly likely that my involvement in that project will be a major factor in my getting a job at some point.
Amusing bits of conference swag:
There were some very interesting talks (and the usual raft of ones that were less so--par for the course), including
During the conference, my advisor Padhraic and I talked about my schedule for finishing up and getting the heck out of here. We're both pretty optimistic that I can do this within a year, given the results that I already have. It would be nice to be done already, of course, but it's good to have external confirmation that my hopes are not unfounded. :)
Random side note: my (Irish) advisor wanted to make a point of going to one of the Irish bars for which Chicago is apparently known. I'm not sure if the person that directed him to the one we went to did so for a joke. . .but while I cannot, sadly, currently recall its name, I will likely forever think of it as The Irish Tiki Bar. Perhaps it looked more Irish inside, but I swear that outside it had tiki torches, umbrellas, and resin chairs; it would have looked right at home in Laguna Beach. And it had Irish quesadillas on the menu. :>
Something I'm thinking about regarding future job directions: of the different industrial research labs that might be interested in hiring me, I've been noticing that some of them publish and some of them don't (much or at all). Microsoft Research, AT&T Labs, Yahoo! Research all publish in my field, but Google and Amazon hardly do. Offhand, this seems to be a pretty clear-cut difference between those companies that have a designated research arm and those that don't, so this isn't that surprising, but if I want to keep my options open for going into academia later, I'll want to keep this in mind.
While I was at KDD, one of my fellow graduate students was working on a joint project called the KDD Challenge (different KDD) whose due date was, at the time, the Friday after the conference, at which time I'd still be out of town. As many of the models were being run on my personal home computer (by me via VNC when I could get a network connection, and by Jon when I couldn't), I spent a fair bit of time on the phone talking him through things, troubleshooting, and generally trying to figure things out as best I could. (\begin{sarcasm} [INSERT DEITY HERE], I love doing phone tech support.\end{sarcasm} On the other hand, as I said to Jon, I recognize that if he hadn't been there to work on it while I was out of town, we couldn't have accomplished nearly so much.)
Anyway, eventually we managed to get things sorted out and finished, and two days ago (the revised deadline) we submitted our results. Earlier that day we got to present some preliminary performance guesses to one of the project sponsors, who was quite pleased. (She's been after us for years to include stuff that had to do with temporal models; I've never seen her so happy as she was Wednesday afternoon. :> )
While I was in Chicago, Megan did a fair bit of gallivanting around with Corwin, mostly to see some of her relatives. Due to some *ahem* disputations with her dad on the subject of the advisability of his driving to Seattle (a) with both a 7-month-old and and an untrained rambunctious dog who likes to lap dive and (b) on almost no sleep (particularly scary in someone with sleep apnea) the Seattle sidetrip was postponed until my return. Unfortunately, as a result, we didn't get to see my of my friends or family in the area, but we did have a nice relaxing time hanging out with some of her relatives up there.
At our farewell party, graciously hosted by Megan's family's friends, we also got to see a bunch of our friends, some of which we haven't seen in a very long time, including![[livejournal.com profile]](https://www.dreamwidth.org/img/external/lj-userinfo.gif)
naudiz (who, perhaps surprisingly, managed to not actually burst into flame at being in the direct presence of The Cuteness of Corwin), pjack, miladycarol, and of course my sister red_frog (who mentioned on her own LJ that she's not convinced that Corwin is actually a real baby, on the theory that real babies aren't actually that cute--I think I'm sensing a theme here :> ).
As a final note, it appears that Corwin travels well: in the car he generally responds well to singing, and on the airplane he's just fine as long as you're prepared to bounce him up and down about 3 million times. :)
Most readers of this here blog are probably aware that Megan and I recently took the hit show "Corwin!" on its first ever road trip, with several dates in and around Portland and Seattle. We anticipate lots of cute pictures with Faces Not Previously Appearing In This Film.
Before I left for KDD, I had a chance to catch up with my friend Danyel, and to attend our friend Mara's wedding reception (the timing of the latter having determined the timing of our trip). At the reception, there was a girl, perhaps 11 years old (the daughter of one of Mara's friends, whom I'd first met before she gave birth--oh, how the years fly) who attempted to recapitulate some of the basic techniques of social network studies in about an hour, with some amused guidance from Danyel and I. :)
Some of you may not have known that my mysterious absence of a few days in the middle of the trip was not due to the usual alternate-identity crime(fighting) spree, but to my attendance at a scientific conference in Chicago, called KDD (Knowledge Discovery and Data Mining).
The primary purpose of my attending this conference was to present a paper at one of the workshops which preceded it. (For all 2/3 of you who care, it was "EventRank: A Framework for Ranking Time-Varying Networks".) This is my first publication in a peer-reviewed workshop. (I have several prior publications, but all of them are either not peer-reviewed, or only "published" in the loose sense that my MSc thesis is available in the UBC library--I never did manage to get my MSc advisor to help me get (part of) it actually published. Maybe someday I will--it's a pretty good guess that no one will have bothered to scoop me in the meantime. :P :) )
The presentation of the paper went very well. I didn't have time to give a practice talk, or even to run through it ahead of time, but my advisor and I were told by a number of people that it was the best of the workshop talks, and several people spoke to me over the next few days about the possible applications of our framework to their problems. (I was amused by the fact that my advisor apparently became more impressed with the talk, and the paper, every time he thought about them thereafter.) Having attended a number of the other talks, we think that if we had done the work a few months earlier (i.e., prior to the conference submission deadline), it would have been accepted to the main conference exactly "as is"--ah, well. (Conference publications are, in my field, considered more prestigious than workshop publications.) Anyway, a somewhat expanded version will get published in a journal or another conference soon, I expect.
This will not be a big surprise to anyone, but authors of papers don't necessarily do the basic background research they need to do. (I'm probably guilty of this myself, but I like to think that I'm more forthright about acknowledging my own failures in that regard when they're pointed out to me.) At least one presentation was given by a researcher who seemed to be treasuring his ignorance, however. *sigh* If this happens to me, I hope that I will at least have the grace to admit that there might be something that I missed.
Of course, an important secondary purpose for attending a scientific conference--especially for one such as myself who hopes to be Making a Career Transition in the foreseeable future--is schmoozing. (My friend Danyel has stated in the past that schmoozing is actually the most important reason. I don't entirely agree with him, but he's got a point.) The well-schmoozed included people from various different academic institutions (Stanford, UMass-Amherst, U of Washington, Carnegie Mellon U), several people from Yahoo! including the new head of Yahoo! Research, some folks at Amazon, and someone from Google (who apparently can't hold a conversation with a possible prospective employee without asking an interview question--fortunately I answered it correctly).
I talked to a number of folks about JUNG. People continue to be impressed. It seems increasingly likely that my involvement in that project will be a major factor in my getting a job at some point.
Amusing bits of conference swag:
- Amazon was giving out T-shirts that say "We have data. Can you read minds?" on the front, and with three ROC curves of increasing quality on the back (labelled "their ROC", "our ROC", "our ROC with you"). Most pleasantly geeky in a machine learning/data mining sort of way. (I can explain the joke if you like, but it's technical enough that unless you already get the joke you may not care.)
- Yahoo! passed out folders with self-promotional material in them. On the front, rather than the full Yahoo! name, they were labelled with "Y !" I was greatly tickled at my thought that perhaps this should be read "Why not" (it's a C/C++/Java joke), but none of the Yahoo! employees that I asked seemed to know whether this had been intended.
There were some very interesting talks (and the usual raft of ones that were less so--par for the course), including
- Jennifer Neville's talk on time-series prediction using network data, which I'll now need to cite as related work when the link prediction work gets published
- Prabhakar Raghavan's work (with Jon Kleinberg) on query incentive networks
- work by Jon Kleinberg, Christos Faloutsos, and a grad student whose name I can never remember (Jure something?) on graph generative models ("forest fire")
- this model--which accounts for the somewhat non-intuitive observation that for large social networks, graph diameters tend to shrink over time, and graph densities tend to increase--highlights my general belief that in doing social network analysis one needs to distinguish between the events that induce relationships (or vice versa) and the relationships themselves. (The total number of people with whom I will collaborate will of course continue to increase over time. The number of people with whom I have active working relationships, however, is bounded.)
- I was bemused to note that the presentation of this paper did not seem to touch on what would seem a most obvious point, to wit, the fact that their model seems to have a compelling story, i.e., it resonates with my intuition about how new relationships are formed when a person is added to a social network. They discussed the mathematical models and gave compelling evidence that it does a good job, but nothing about why this model might actually make sense in the real world. Odd.
- this model--which accounts for the somewhat non-intuitive observation that for large social networks, graph diameters tend to shrink over time, and graph densities tend to increase--highlights my general belief that in doing social network analysis one needs to distinguish between the events that induce relationships (or vice versa) and the relationships themselves. (The total number of people with whom I will collaborate will of course continue to increase over time. The number of people with whom I have active working relationships, however, is bounded.)
During the conference, my advisor Padhraic and I talked about my schedule for finishing up and getting the heck out of here. We're both pretty optimistic that I can do this within a year, given the results that I already have. It would be nice to be done already, of course, but it's good to have external confirmation that my hopes are not unfounded. :)
Random side note: my (Irish) advisor wanted to make a point of going to one of the Irish bars for which Chicago is apparently known. I'm not sure if the person that directed him to the one we went to did so for a joke. . .but while I cannot, sadly, currently recall its name, I will likely forever think of it as The Irish Tiki Bar. Perhaps it looked more Irish inside, but I swear that outside it had tiki torches, umbrellas, and resin chairs; it would have looked right at home in Laguna Beach. And it had Irish quesadillas on the menu. :>
Something I'm thinking about regarding future job directions: of the different industrial research labs that might be interested in hiring me, I've been noticing that some of them publish and some of them don't (much or at all). Microsoft Research, AT&T Labs, Yahoo! Research all publish in my field, but Google and Amazon hardly do. Offhand, this seems to be a pretty clear-cut difference between those companies that have a designated research arm and those that don't, so this isn't that surprising, but if I want to keep my options open for going into academia later, I'll want to keep this in mind.
While I was at KDD, one of my fellow graduate students was working on a joint project called the KDD Challenge (different KDD) whose due date was, at the time, the Friday after the conference, at which time I'd still be out of town. As many of the models were being run on my personal home computer (by me via VNC when I could get a network connection, and by Jon when I couldn't), I spent a fair bit of time on the phone talking him through things, troubleshooting, and generally trying to figure things out as best I could. (\begin{sarcasm} [INSERT DEITY HERE], I love doing phone tech support.\end{sarcasm} On the other hand, as I said to Jon, I recognize that if he hadn't been there to work on it while I was out of town, we couldn't have accomplished nearly so much.)
Anyway, eventually we managed to get things sorted out and finished, and two days ago (the revised deadline) we submitted our results. Earlier that day we got to present some preliminary performance guesses to one of the project sponsors, who was quite pleased. (She's been after us for years to include stuff that had to do with temporal models; I've never seen her so happy as she was Wednesday afternoon. :> )
While I was in Chicago, Megan did a fair bit of gallivanting around with Corwin, mostly to see some of her relatives. Due to some *ahem* disputations with her dad on the subject of the advisability of his driving to Seattle (a) with both a 7-month-old and and an untrained rambunctious dog who likes to lap dive and (b) on almost no sleep (particularly scary in someone with sleep apnea) the Seattle sidetrip was postponed until my return. Unfortunately, as a result, we didn't get to see my of my friends or family in the area, but we did have a nice relaxing time hanging out with some of her relatives up there.
At our farewell party, graciously hosted by Megan's family's friends, we also got to see a bunch of our friends, some of which we haven't seen in a very long time, including
naudiz (who, perhaps surprisingly, managed to not actually burst into flame at being in the direct presence of The Cuteness of Corwin), pjack, miladycarol, and of course my sister red_frog (who mentioned on her own LJ that she's not convinced that Corwin is actually a real baby, on the theory that real babies aren't actually that cute--I think I'm sensing a theme here :> ).As a final note, it appears that Corwin travels well: in the car he generally responds well to singing, and on the airplane he's just fine as long as you're prepared to bounce him up and down about 3 million times. :)