eric brochu | haiku factory

thesis page count, originally uploaded by Mister Wind-Up Bird.

Holy crap. It’s 2:13AM on a Friday night, I’ve got Mogwai playing on the stereo, and I think I might actually kind of be done writing my thesis. I feel… odd.

(Well, done until Nando and my committee read it and tell me to go revise it all, but still.)

I’m a little late in posting this, but I managed to get accepted to the 2010 Symposium on Computer Animation with my paper, “A Bayesian Interactive Optimization Approach to Procedural Animation Design”! I’ve added the pre-print PDF to my publications page, and there’s even a video, which explains the whole thing in under five minutes (with no math). So lest anyone question what a Machine Learning person is doing trying to get into a computer animation conference, behold!

I’m pretty happy about getting this accepted for a couple of reasons. First, it’s a big part of my thesis, so getting it accepted to a good venue is a major step toward finally graduating. And second, it means a free trip to Madrid to present it! Hopefully I’ll be mostly finished writing my thesis by that time (early July), so I’ve already booked a bit of an extended tour that takes me to Paris for a few days before the conference and lets me stay in Madrid for a couple of days afterward. My brother-slash-coauthor Tyson gets to go, too, so that should be fun. Especially as he actually speaks French, so he can deal with any potentially snotty Parisians for me.

After that, it’s off to Australia to visit Janelle. I’ll be flying in and out of Brisbane, which is cool, but we’ll also be spending a good chunk of time in Melbourne, which I’m stoked about. From all accounts it’s a really cool city. I’m imagining it as kind of a mix of the best aspects of Vancouver, Toronto and Montreal. Will I be disappointed? Don’t worry, I’ll let you know.

I’m also hoping to talk to people in Melbourne in the research community (for start-ups in particular, but in general, too), so if you happen to know anybody, or know anybody who knows anybody, or you are that anybody, you know what to do.

phantasy of phinishng a phd, originally uploaded by Mister Wind-Up Bird.

So for those keeping track at home, I took leave of my position at Worio last month to go back to work on my PhD full-time. I’d been either a non-student or part-time for the last two years, which seemed kind of incredible to me. How the hell did a six-month leave turn into two years? I guess by extending it and then getting sponsored to be a research intern, is how, but that answer seems too literal and lacking the existential resonance I was hoping for. So I’ll just say that the PhD is simultaneously the best and worst thing I’ve ever done. I’m glad to be back and gladder that the end is finally in sight.

But for now, I’m fully immersed in the full-time grad student lifestyle again. It all came back easily, as I immediately had to dive into submitting a paper to a fairly high-profile conference. Which means the past month has been the tick-tock of an approaching paper deadline with far more to be done than time to do it. Days with the sickening feeling of knowing you’re procrastinating and feeling powerless to stop it. Other days enjoyably lost in elaborate code, listening to Four Tet and Aphex Twin. The sudden happy click when a once-obscure paper starts to give up its secrets and make sense, or the rush of getting a really clever idea for solving a problem. And then biking home from the lab, exhausted, in the early morning hours, and passing through Kitsilano, which is still and quiet and smells like flowers and marijuana and the ocean.

All that, and I kind of think the paper I submitted was not-so-good. The code stayed buggy, the experiments unconvincing and the writing something less than a model of clarity. Oh, well, the “good” news is, there’s always another conference deadline coming up soon!

Until I’m finally done, and then there isn’t.

So I’ve officially started writing my PhD thesis. I’ve downloaded the latex template and written an outline and everything.

My typical thesis-writing day goes like this:

The lovely Janelle has been visiting, so I’ve been neglecting the ol’ blog in favour of spending my free time doing things with her, like eating calf brains at Boneta (verdict: delicious) and watching Project Runway (verdict: hell of a lot better than I expected). But she’s in the spare room sewing and listening to Kevin Smith’s podcast at the moment, so I thought I’d give my as-promised report on how the thesis is going.

It’s been an interesting transition back to full-time PhD-dom. While working, I was much more focussed on productivity than creativity, meaning I went to Yaletown around 9 every morning and left around 6 or 7 every night, and worked on mostly fairly-immediate new features and bug fixes. When I did research, it was to find the solution to a problem, not to investigate something novel. While perhaps not entirely “creatively fulfilling”, I do get a lot of satisfaction just from the feeling of accomplishing and producing something pretty much every day.

The grad student lifestyle is a big a shift from that. For one thing, it’s a lot less structured. I am still working with Worio a couple of days a week, which forces me to manage my time a bit, but I generally work at different places and different times and decide day-to-day what needs to be done. This is not great for short-term productivity, but I find it very important for creativity. As an academic researcher in Machine Learning, there’s not usually a lot of payoff in incremental improvements and fixes — you need to take more chances and do something new. Which is why I spent a chunk of January working on a conference paper that ended up not being submitted. Most of the rest, I spent reading a couple of books and a stack of papers, some of which were interesting and some of which weren’t. I’m much more of an applied-science guy than a theorist, and much more of a doer than a reader, so the upshot is that I sometimes have to force myself to sit down and do the reading instead of diving in and doing the coding and experiments. But I think it paid off. I literally had one of those in-the-shower moments, where you have to get out of the shower, grab a notebook and write down a multi-page detailed description of your idea, complete with sketches of figures. It all felt very scientist-y. If it works, this is a publication and a chapter of my thesis. If it doesn’t, it’s a big chunk of lost time. I’ll know in a few weeks.