Study one thing well, then branch out
Since finishing my doctorate and starting a non-research oriented position, continuing the work from my thesis has been difficult. This is because I don't have much time to devote to it. Much of my research time consists of me ruminating on subjects while driving.
When I do have time to read, I've found it much more fruitful to branch out and study some other topics besides computer science. Two of those have been physics and economics.
Even reading introductory texts on the subjects have provided me with a wealth of ideas. My doctoral work got me very deep into a single thing, even though I tried to diversify in some ways. Branching out to other subjects just to get the big ideas has helped "connect the dots" in ways I had not considered (showing me that I wasn't as diverse in my ways of thinking as I thought!). I'm going to be so bold as to say I doubt I would connect those dots were it not for the fact that I read something in both physics and economics.
What I draw from this is that it is helpful to study one thing in depth, but then branch out and study other things. This is not a revelation, but I think it is important to note.
Reflecting on my thesis
On Monday I successfully defended my PhD thesis entitled Structuring data via behavioural synthesis. My thesis looked at ways to derive the structure of data given the behaviour of a system. That is, by studying the actions of a system you can determine the structure of the data objects it uses.
It's been a long, hard road getting to the this point. I started my graduate work six and half years ago (January 2002) and started out working in bioinformatics. I then got the itch to work with programming languages and the work finally led me to examine the dynamic elements of programs. In particular, I got interested in context-oriented programming.
Only work from the last two years actually made it into the thesis. While that may seem to indicate the previous four and a half years was a waste, nothing could be further from the truth. The steps leading to the work I did were important because I'm pretty sure I wouldn't have had the background I did without making various mistakes and missteps along the way. Ideas evolve and rarely spring into being, and I think some of the ideas I examined were greatly helped by making the mistakes I did.
Richard Gabriel was my external examiner. I cite him in my thesis and his work motivated me to look at what I did. Needless to say, I was a little nervous. However, he asked some great questions and got me thinking about where to take the work a bit more. I will certainly be exploring different avenues with it.
I want to thank all my friends who have been supportive of me over the years. You have been a big help in keeping me sane. In particular, I want to thank Ren, Greg, Ian, Adam and Les.
I don't say I'm proud of very much, but I'm proud of my thesis. I put a lot of work into it and went through a lot of stress to get it written. Even if no one else ever reads it, I'll be happy with it.
When I have the revisions complete, I will post a link to the thesis. (Some people may like the first two chapters, actually.) I'd rather not provide a version that requires changes. I'll also work on posting it in HTML rather than PDF, since I'm sure few people will want to print it.
Thesis writing is tough, but rewarding work
For the last month, I've basically been locked away writing my dissertation. Here are some things I have noticed.
- Your thoughts tend to get very focused. My thesis is the adaptation of some previous papers I have written, but with a lot more formalism to lay out the idea more clearly. Papers often have limited space, which means you can rely a bit more on intuition. I still have to be concise, but more rigourous. This means careful consideration of the argument.
- Consequently, figuring out what to write about next can be time-consuming. I try to avoid jumping between ideas in paragraphs too much to prevent discontinuity. Further to that, I don't like excessive sectioning to delineate ideas. I try to put a narrative together and give it some flow without forcing it.
- It pays incredible dividends to sit down with a piece of paper and sketch out the main points of your argument before trying to write anything. Outlines are invaluable.
- Having a reasonably diverse vocabulary makes your first draft come out a little faster. I don't sit for a long time trying to figure out how to say something. Most of my time is spent figuring out what I should really be saying.
- I like writing. It's difficult, but rewarding. (Most of the above is really a side-effect of sitting down to write.)
- I really, really appreciate LaTeX. I prefer TextMate to Emacs for the actual editing.
- Version control is something that should be mandatory for any writing/text editing software. Failing that, it should be in the operating system. Time Machine saved me untold hours of work from an incident that erased a section of a chapter I really didn't enjoy writing. I didn't notice the section was gone until hours later (and I was being careful! Or so I thought). With Time Machine, I restored it in under five minutes.
Beauty, theories and equations
I really enjoyed the following talk by Murray Gell-Mann.
I think what he says applies outside of particle physics as well, but there are some elements ripe for discussion in the presentation. The first one that struck me was how the equations got simpler and more beautiful seemingly by coming up with better notation. Unfortunately, he didn't get into the details, but I don't outright disagree with the idea as it was presented.
The mismeasurement of science
Insightful commentary? Astute observations? Curmudgeon with a point? A little bit of everything, I think. (See an accompanying blog entry as well.)
Worthwhile readings on science
References can take you in weird directions.
Last night I was reading a transcribed panel discussion from a conference in 1986 about language and problem solving, which lead me to the paper Strong Inference [subscription required] by John R. Platt, published in 1964, that talks about how important it is to come up with testable hypotheses in a rigourous fashion to prevent you from blindly groping around the search space related to your work.
Platt's paper refers to a paper by T.C. Chamberlin, The Method of Multiple Working Hypotheses. Although the language is a little baroque (did I mention it was first published in 1890?) it is well worth the read. He argues that we must avoid the desire to get attached to our ideas by generating many of them so that we do not fall into the trap of the "ruling theory," that is, the theory we bend the facts to fit so that the theory is correct. A single hypothesis is a step in the right direction, but it has the tendency to turn into a ruling theory. By forcing yourself to entertain a variety of possibilities, you widen your understanding and avoid single-mindedness.
Platt's and Chamberlin's papers have little to nothing to do directly with what I'm interested in, but they were worthwhile reads to keep me on track and focus my thoughts. If you're science-minded, I recommend checking them out.
Evaluating research
Daniel Lemire linked to a fascinating paper that discusses how research is evaluated and the associated hidden costs.
For those interested in research, I recommend reading it. The only downside of reading it is that it has fueled some of my cynicism regarding how I see research being conducted. Admittedly, I'm not involved in any evaluation committee or do any peer review, but I do see the results of it. The "salami" model of publishing disturbs me because it seems, in some way, dishonest. It's something I can't quite put my finger on.
What I didn't take away from the paper is that peer review or evaluations are bad and should be abolished; it's regarding them as sacrosanct in their current form that is the problem. I've never been one to take much as a given, so I'm all for trying something different.
Technorati Tags: peer review, academics
The ILC post-mortum
Overall, ILC 2007 struck me as a great success. Nick Levine should be upgraded to demi-god status for all the work he put into it. Almost everything was precise and professional.
Of course, there's a snag: internet access. In short, it sucked. I don't really blame Nick for this; I don't think Cambridge has a very good set up. More than a few conference delegates suffered the horrors of the MemorialCourt wireless network. People staying in conference-supplied rooms had no internet access from their rooms: they had to go to the common area (in a separate building) and cross their fingers. Signal strength was random and two people with the same kind of machine sitting right next to each other with the same setup would get totally different results.
Oddly enough, the lack of net access in the presentation room was probably a good thing as it prevented mass distraction. It would have been nice to get it out in the lobby, though, so people could go download some of the things being shown and try them out while the desire was still fresh.
The food was excellent. We were well fed with a decent variety of dishes provided. Absolutely no complaints there from me.
That's enough for logistics. I want to say a few words about the presentations themselves. For the most part, I was disappointed. Disregarding content, I thought that most people demonstrated a lack of care in presenting their ideas. I made a conscious effort to present my work as clearly as possible. I failed in some respects, but I sincerely hope that some others didn't put forth as much effort as I did to arrive at the results I saw. (I was told that my talk was good by a few people and that they really thought this. Thus, they felt better that they didn't have to lie when they would inevitably come up and say my talk was good.) Although I dismissed official talks a bit earlier, I do consider it important to be able to convey your ideas. A lot of what I saw didn't fall into this category.
Yeah, it's wishful thinking to want better presentations when poor ones are universal. I say it with sincerity and without contempt, though. I really would like to see a bit more forethought when it comes to showing others your ideas.
Speaking of which, my MacBook Pro was flawless when it came to hooking up to the projector. Most people seemed to get by without much trouble, but then again, I wasn't a session chair.
There was a high proportion of Apple hardware being used. It really seemed as though non-Apple laptops were the minority. I have no idea if this says anything, but I certainly noticed it.
A concern I had that was evident in more than a few talks was work that had pragmatism mixed with lots of enthusiasm, but hasn't seen it followed through to the end yet. This is the "80% problem" I mentioned earlier: projects get a good start and are interesting, but never seem to finish the tough part, which is the last 20 percent.
I can think of three works that exhibited this: the translation of Java libraries to Lisp, XMLisp, and compilation of list comprehension. XMLisp looks really neat and works to some degree, but it's not a complete XML implementation. This is something I feel should be seen to completion and it would be a shame to see only part of it working; being able to write XML directly could be incredibly useful. But in order for it to be taken seriously, it has to be a full implementation. cl-ppcre wouldn't be so great if it didn't do all of Perl regular expressions and I commend Edi for "finishing" it, as it were.
The list comprehension compilation is another example. Here's a really useful macro, but it needs to be able to handle things like functional generators and be crafted so that it's easy to use. It also needs to do more serious compilation by doing some symbolic analysis. This work should be seen through to a more robust and complete implementation and not just one that scratches an itch.
The Java libraries paper shows that some tasks can be difficult to do correctly (not that this wasn't already known). Lots of people said that it's amazing what has been done, but it's not to the point where it's as automatic as the paper seems to suggest (and the conclusion of the paper addresses this point, which is odd). I was glad to see the paper noting some of the shortcomings of the work and I hope it continues, but from what people told me about the talk, the actual approach still needs work.
I'm not immune to this. I didn't try to show my work as some awesome solution to a difficult problem, but reflecting on it a bit, I think I didn't do a great job of following up on some themes, mainly practicality and purpose. This may be a case of excessive self-criticism, but I have this nagging feeling that I need to validate my ideas by applying them to working projects. (I had more than a few of the "How can I use this?" kind of questions.)
This is always a problem with the framework/idea kind of work, so during the conference I set to work on making use of the dynamic ADT idea that I pitched and I think I came up with something kind of fun and even, dare I say, practical. Imagine that! Rather than go way off track, I'll post something about that later.
Anyway, enough of the philosophical digression. The point is that there is too much important work being done in Lisp that isn't finished and the people working on it get distracted by other shiny objects and forget about it. That would be unfortunate and I don't want to see that happen.
I definitely want to go to the next ILC and will happily get the money together to pay for an ALU membership. This conference was a great time. I talked with some really good people, got to meet some really good people and put some faces to names. It also makes me want to be more active in the Lisp community. Despite my reservations and concerns, I think there's a good thing going with this.