July 16, 2011

PhD

Tags: Memories, University

After honours I was offered a government scholarship to study for a PhD. This was the standard way of financing postgrad study, all the other PhD students in the UWA CS department had the same scholarship. I started the PhD in January 1996, but by September 1997 I had deferred the scholarship, started looking for a job and unofficially quit my studies. The official end came 6 months later when the scholarship was rescinded and a letter arrived saying “don’t ever come back” (I paraphrase). I don’t regret that year and a half. I learnt a great deal about logic, a little bit about programming and substantially more about the world outside of computer science - all paid for by the Australian taxpayer (thanks guys!).

I was lost at the start of my postgrad work. I knew I wanted to do a PhD, but not what I wanted to study. This was a bad sign. I knew what I didn’t want to do: neural networks, my Honours thesis used them and they seemed like statistics-based black boxes. I wanted to see what was happening (if only to assure myself it was working in the manner I expected). There was a new lecturer, Kym, whose research focused on logic. I asked him to supervise me and he suggested my topic Using Non-monotonic Logic to extend Inductive Logic Programming. I thought that sounded interesting and started going through his suggested readings.

Read, read, read. That was almost all I did for the first six months. I found it hard going. Maintaining concentration throughout some of the longer articles was difficult. Since I didn’t understand my thesis completely, I read a lot of papers that weren’t relevant. Especially papers on complexity theory as they sometimes had non-monotonic and logic in the title. They were particularly dense and not useful to me. At this stage the closest I came to coding for my project was working on my BibTeX bibliography. After six months I gave a talk to the department and was “locked in” to my project.

Most of the postgrads worked on the ground floor of the CS building, grouped by supervisor (and thus roughly by area of research) and placed into one of four labs. The Robotics and Computer Vision lab was the largest and best funded. They had the best computers and various cool robots. I remember at one point they got a server with 768MB of RAM - it was the most powerful computer in the department (in 1996). The next largest lab was the AI lab. It was just a set of desks and dividers creating little cubicles. The people in there didn’t talk much to other people - or at least not to me. Their door was usually shut and no sound emanated from within. They seemed very serious. The Computer Music lab was the opposite, it had just two people, but they were both very friendly and welcoming. Then there was my lab - the Miscellaneous Lab. Officially it was called Room 1.14 or something like that. We renamed it the Misc lab. I suppose by topic I should have been in the AI lab, but I preferred being in Misc - it was a more fun room. We were all the students of lecturers who didn’t have the big research grants lavished on the CompVis and AI lecturers. There were five official occupants of the lab, but one preferred to study at home (smart move). The rest were Luigi, Rowena, Kamran and myself. We got on pretty well as I recall and talked together often.

As my supervisor was new to the department and didn’t have any grants yet, I got a PC from one of the undergrad labs. I also had a 21 inch monitor, with which I was immensely impressed - it was the biggest screen I had ever used. The PC was a Pentium 90MhZ and it ran Linux, which I installed myself. I could have used my departmental grant to upgrade it, but I wanted to save that money. Each student was allocated a few thousand dollars by the department to spend on equipment or travel. Most postgrads used theirs on airfare to go to conferences. I only ever used mine to buy a published conference proceedings (about $40 in total I think). I never went to a big conference. I wanted to have something completed first. However, I did go to the departmental conference twice. They were good fun and not too serious. One year a student gave a talk on ice hockey. They appeared to be for the postgrads to practice presentations (and to boost the departments publications count).

After reading for what felt like forever, I needed to code again. There were 3 projects of note during the last year of my studies, although only one was related to my research. I implemented a learning algorithm detailed in a paper that was similar to what I wanted to achieve. I then improved it a little with some ideas from another paper and tried running it on a large dataset that naturally lent itself to exceptions in the rules learnt - the KRK dataset. A list of all 28056 possible chessboard positions for two kings and a rook. My program would then try to learn the rules for determining whether a position is legal. I wrote the system in Prolog and it took many days to run to completion. I would usually start it on Friday afternoon before heading out and come back on Monday to see if it had finished. Often my computer would crash as it ran out of memory, but eventually I got it to work and output the correct answer. This was the only notable achievement of my PhD.

I then planned a new learning system. However, instead I ended up designing a logic resolver - something like Prolog. I never wrote much code, but spent time planning it. It would have been of little benefit to my research (Prolog already did the job acceptably). Also the design was poor. This was not the worst code I wrote while studying. That dishonour goes to a library I wrote to add functional coding to C (lazy evaluation, partially evaluated functions, etc). It was buggy, leaked memory like a sieve and converted everything to void *. It was a bad idea from the start. I didn’t use it much, but Luigi did for a while before abandoning it as broken. I hope I didn’t set the course of his PhD back too much (he completed). Despite all this non-thesis coding most of my time was spent not doing much. Counting programming, research reading, working on an unfinished literature review or other documents, I probably only averaged a few hours per week of useful work over the course. A large drop from my honours year. The rest of the time was spent exercising, socialising, going to lectures in other departments for fun (first year philosophy was a favourite), teaching and procrastinating.

Teaching was a surprisingly large timesink. All the PhD students complained, but did it anyway as it was one of the few ways we could earn a little cash on top of our scholarships. It was decent money too, around $20/hour, but we weren’t supposed to do more than a few hours per week. A rule honoured more in the breach. I was a tutor for 3 semesters of the introductory CS course as well as a semester of databases and another of logic. It was fairly easy work, making me wonder why I had any difficulty when I did them. I could do most of the first-year assignments at the speed of typing. My tutorials could be given without preparation, though this meant I sometimes covered topics outside the assigned coursework. I once explained Garbage Collection to a completely lost group of first-years - then had to spend a few minutes assuring them they wouldn’t be tested on it. I tried to help the students as much as possible - I genuinely enjoyed teaching. The majority of my time was spent on a few students in each class. Like me when I was a student, most students did not ask for help regardless of whether they needed it or not. Those that did tended to ask a lot, which was fine unless they became aggressive after not understanding (one student) or relied on me too heavily to explain things. I remember one student (called Doug I think), who refused to lean anything not taught to him in a one-on-one session with a tutor (often me), despite clearly being smart enough to work it out himself. One attractive student gave me a box of chocolates for helping her before an exam. I was on the receiving end of a few good-natured barbs for that.

The downside of teaching was the marking. Every few weeks there was another assignment to be marked, and the markers soon wished more students asked more questions! We would get a pile of 30 or so assignments and have a week to mark them with constructive comments. It could take a long time. Luckily although we had 30 papers, there were normally only about 6 assignments. All the rest were copies. The most obvious copies were flagged as such, but usually nothing would happen as the students knew how to talk their way out of it and it just was not worth a fight for the lecturers. Most of the copies changed the comments, variable names and maybe modified the code a little. Enough to claim that they had worked with someone else (which was allowed), but had completed it themselves. Maybe that is true, but I think plagiarism was rife. At least the copies were quick to mark. It was the ones that didn’t copy and made big mistakes that took time as we had to work out where they had gone wrong so part marks could be assigned. This usually involved trying to discover the tortured logic the student was following. Thus I discovered the secret to good marks: make your teacher’s job easy!

Socially, my PhD was split into two halves. In the first half I hung out with the other postgrad students. We had lunch together and occasional trips to the pub. After work I generally went home (I lived by myself). There were regular poker games, but I never got involved with those - there were too many good players. I didn’t have a computer at home. I had sold it when I started the PhD as I had 24-hour access to the labs, which were only 10 minutes walk away. I spent my time in contemplation, reading, listening to music or the radio (I didn’t have a TV). Most weeks I would go to the local pub by myself for their local music gigs. I bought just a single beer, listened to the band, and then left. I referred to it as being social by osmosis. Outside work hours I spent my time alone. I got to know the radio timetable well and gained an appreciation for Aussie Rules Football through the broadcast commentary. I remember Christmas 1996 fell on a Wednesday. The department was locked and I couldn’t get in, so I spent my time at home. I went out on the Monday to buy food and didn’t see or speak to anyone until that Friday when I went food shopping again.

The second half of the PhD was completely different. I regularly listened to a house and techno radio show called Strictly Rhythm on RTRfm. The guy who presented the show also ran a record store and nights at local clubs. I started to frequent his store and attend his clubnights. I met many people through these events and became a peripheral member of the local “scene”. I went out to clubs at least once a week. This wasn’t too expensive because I was always on the guestlist and usually got a free drink or two. I even DJ’ed a few times. The details of this time deserve a post of their own. Suffice to say that my research suffered. I also moved into shared housing. My lease ran out and I needed to find somewhere cheaper. Leigh, another postgrad needed some extra cash, so I moved into his closet. In my defence it was a very big closet and had a door to the back garden. I think it used to be a patio that was walled in. I was only there with Leigh and Dougal for a few months before I moved into another sharehouse with Leigh and two others. Once again the damage was done, I had a social life at home. I even met a girlfriend through one of Leigh’s friend’s parties.

By mid 1997 it had been a few months since I had done anything of note for my postgrad. My weekly meetings with Kym were awkward affairs. Eventually he said I need to get motivated and moving again. He gave me an ultimatum, to write a paper reviewing my research area for an upcoming small conference. I thought that was a good idea and started work. After a few days, I decided to quit the whole PhD. I found myself wondering why I was writing this paper. I didn’t enjoy it, I’d rather be coding or listening to music or going out. Academic theoretical research is just not for me. I prefer to be building something - more engineer than scientist. When I began the course I wanted a PhD, but I wasn’t sure why. It was not because I enjoyed the process or thought I would enjoy being a lecturer or researcher. It was instead just another goal to reach. A mark of success. I decided to quit my PhD and get a job. I let Kym know the next day - he didn’t seem surprised, but suggested I suspend for six months instead. The Head of Department called me into his office and told me that if I suspended I would never come back. I agreed, it was unlikely I would return. He said that in that case it was better to go sooner rather than try to carry on for a few more years. He wished me luck and I handed in my forms the next day, collected my stuff, burnt a CD of my data. That weekend I had some friends around to watch me barbecue some notes and papers as a symbolic end to the process.