Monday, August 22, 2005

Rabin's PhD

I first met Rabin Ezra some time in the second half of the 1980s. I taught the Computer Graphics course, a final year course at Queen Mary and Westfield College, Department of Computer Science. At that time the lab sessions were in an area of the college called ‘Stern Hall’. I had a lab class in there, and a young student approached me, someone I’d not seen before. He explained that he was an undergraduate in the Electrical Engineering Department and wanted to study the Computer Graphics course in the next academic year. After checking that he had the required knowledge of computer programming, I agreed. He seemed to be a very interesting student, and very interested in computer graphics, with also a good prior background that would enable him to do the course.

During the course itself I had several after-class discussions with him, and it was during one of these that the issue of his staying on to do a PhD arose. He was very interested in hardware for real-time computer graphics. Also I had read a very interesting paper by some people at Sun Microsystems that provided a very efficient method for the rendering of curves and surface that would be amenable to hardware implementation. Rabin became interested in this and we agreed upon this as his topic. He started the PhD probably in 1988.

He made rapid progress and actually soon had a hardware design for the main part of the surface generation graphics hardware. He passed his first and second year vivas without trouble. But then he did run into trouble. It seemed to be a trivial problem. Suppose you are given n points in 3D space. You need to decide if all these points are approximately on a plane. Mathematically this is not difficult. Take any three of the points that are not themselves on a straight line, and from these find the equation of the plane that passes through them. Then for each of the remaining points find their distance from this plane, and if the maximum distance of all these points from the plane is less than a certain fixed and given tolerance, then we can say that ‘approximately’ (how approximate depends on the chosen tolerance) the points are flat, i.e., are on a plane. Mathematically this is trivial. Transforming it into an efficient and accurate computer program is less trivial. Fitting it into a hardware design architecture so that it doesn’t become a major bottleneck is very difficult. This is what occupied Rabin for the remainder of his PhD. Finding some workable solution to this problem took about 2 years.

In January 1991 I took sabbatical leave from QMW to take a Visiting Professorship at University of California Berkeley in the Computer Science Division of Electrical Engineering and Computer Science. There I also taught the Foundations of Computer Graphics course, and continued with my own research. I was in frequent email correspondence with Rabin about the progress of his PhD, at that time he was still stuck on the ‘points on a plane’ problem described above. Having such a problem is not a ‘bad’ thing for a PhD because it leads to research avenues far from those envisaged at the start of the PhD research, and we certainly did come across a number of very interesting papers and lines of investigation that we would not have found but for this problem. Meanwhile it was during this period that Rabin became fascinated also with the ‘systems’ aspects of computer science. The Computer Science Department at QMW had been one of the first (or possibly the first) to bring the Unix system to the UK, and there were a number of experts there in this operating system. Moreover there was an active ‘systems group’ of people who continually investigated aspects of this system and others, and who became very technically proficient in these systems. They became ‘gurus’. They would (following the Computer Science at QMW tradition) work long hours into the night doing ‘system’ things, learning, trying, testing, debugging, implementing, solving problems: became a breed apart. Rabin became part of this group, transformed into a guru. This also had a detrimental effect on his PhD progress – no doubt exacerbated by my physical absence.

I was offered additional semesters (this time paid) by Berkeley, and therefore took unpaid leave from QMW in January 1992 until the summer. (I nevertheless had taught all my QMW classes compressed into one term, in October-December 1990 for my 1991 sabbatical and in October-December 1991 for my 1992 unpaid leave – such had become the nature of sabbaticals). My time at Berkeley had been so productive and enjoyable that I was very happy to be going back there for a second time. (I have to say also that teaching at Berkeley was far more rewarding than teaching at QMW at that time, but I think that this is a cultural difference relating to differences in students between the US and the UK. In the UK at the end of a lecture the students who would come to speak to me – in the main – Rabin being an exception – would say “I didn’t understand anything” “This course is very hard” “Is this going to come up in the exam?” and so on. At Berkeley, at the end of the class many students would come to talk to me enthusiastically about the ideas presented in the class, give their own opinions, talk about similar things they were doing in their part-time jobs and so on. It was a great pleasure to talk with them. In this respect Rabin followed more closely the model of the Berkeley student than the QMW student).

During this period Rabin and I had agreed that he would begin writing his PhD. One day shortly after I had returned to Berkeley Rabin sent me an email: “I’ve finished!”. I was very surprised. A PhD thesis normally runs to about 150 pages. Had he written so much, with all the references and all the figures, in such a short time? I asked him to send me the thesis. Well what he had written was fine – but it was about 40 pages in length. He had succinctly stated the problem, and the solution and results, in as terse manner as possible. There was nothing at all wrong with it, except that it really was too terse. So over the next year or so we gradually got more and more flesh put on the bare bones of the thesis. It was somehow against Rabin’s nature to spell things out in detail. “It’s obvious” he would say, "why should I have to write about it?" Well, obvious to him, and to me, but to the examiners, and other readers? – not so obvious. Yes, things have to spelled out, in detail. Where does this equation come from? Why is this statement true? What reference do you have to back up this idea? With some resistance Rabin eventually saw the light. But in a way of course Rabin was right. He had stated the problem and given the solution and results. Really of course the rest is convention. This is how a PhD is supposed to look. To be awarded a PhD he had to learn to follow certain conventions. But … they are only conventions.

He finally submitted in late 1993 or early 1994. The thesis length was still quite short, but acceptable.

On the day of the viva he was extremely nervous, to the point of shaking visibly – Rabin was not one for formally speaking in public, and he would have to do a short presentation in front of the two examiners. I told the examiners about his state. One of whom was Dr Simon Arridge from University College London who said I should tell Rabin not to worry, everything was ok. This did seem to quieten his nerves somewhat, and although still visibly shaking when he did his presentation, he got through it well. In the ensuing viva period he relaxed and engaged in genuine discussion with the examiners, and nothing remarkable happened. He was awarded the PhD.

After the award of the PhD Rabin went to work for Canon Research Europe where he specialised in graphics hardware and software, and especially on the interface between the two. CRE was partially founded by an ex-QMW student, and a number of ex-QMW people eventually worked there.

I kept in touch with Rabin over the years, and in particular used to meet him annually at the SIGGRAPH meeting in the US. I will write more about this at a later time.

Rabin is unforgettable. Although I will never see him physically again, he is still around. Although he passed away before SIGGRAPH 2005, I believe that he was still there, and enjoyed it as usual. He lives on at least through our memories and in our hearts.

Thursday, August 18, 2005

A good friend

Rabin was part of a circle of my friends that met when we were all at Queen Mary College. At the time, in the late 1980's, he was an undergraduate and I was working in the Computer Science department. We discovered a common interest in Role Playing Games which, if you don't know, are a very social activity involving sitting around acting out a story together, often in a fantasy "Lord of the Rings" type environment but also sometimes in futuristic or contemporary settings. The game is really an excuse to get together over some drinks and food and have a good time. Rabin was a well-loved member of our group and his intelligence and wry sense of humour will never be forgotten.

In latter years, as we have got older and ever more busy, it has become more difficult to arrange to meet up as a large group. However Rabin always made the effort to come along, even if it was just a couple of us meeting for a drink, or the occasional barbecues some of us hold. Out of all my friends I think he was the most intelligent, ethical, and dependable. I won't forget him.

Paul Davison

Rabin Ezra

This blog is dedicated to the memory of Rabin Ezra. Here his family, friends and colleagues will post their recollections, memories, thoughts, reflections, so that he will live on, and so that more and more people will come to know him.