by Hinke Osinga, University of Bristol, UK

Gerton Lunter studied both Mathematics and Physics at the University of Groningen in the Netherlands. Already during his school days he was also very much intrigued by computer science and he loved to program. He got an MSc degree in Mathematics in 1994 and continued studying for a PhD in Dynamical Systems under the supervision of Henk Broer and Gert Vegter. "I was always fascinated by the pictures and, in particular, fractals in dynamical systems; isn't it amazing how very complex fractal structures emerge from really simple computer programs?" The particular choice of specializing in dynamical systems was also guided by the fact that dynamical systems is a very broad subject that utilizes many aspects of mathematics and that computer programming plays an important role in the analysis of dynamic behavior.

Gerton Lunter in Cornwall, summer 2003

In December 1999 Gerton obtained a PhD with his thesis Bifurcations in Hamiltonian Systems; Computing Singularities by Gröbner Bases. Even though his thesis contains lots of pretty pictures and complex fractal structures, and despite the fact that the use of Gröbner bases lead to a natural algorithm for deriving normal forms that was "a lot of fun to program", his pleasure in pure fundamental research had dwindled and Gerton opted for a career in industry. "I had started my PhD research believing that dynamical systems was a field with exciting applications that are physically relevant, but during my PhD I sometimes felt that my research was leading nowhere. Moreover, nobody seemed to be interested in what I was doing. I was in need of appreciation for my skills and decided to search for work that was oriented towards specific goals and with very practical applications. I wanted to do something that was of interest to people."

Getting the job

As a child, Gerton had visited the Evoluon in Eindhoven for the special exhibit on science and technology; this exhibition opened in 1966 to celebrate the 75th birthday of the Philips company and ran until 1989. It was here that he learned about Philips Research. Officially called Philips Natuurkundig Laboratorium, this is the biggest company research laboratory in Europe, with a total of (then) 1500 people. It is here where they invented cassettes and CDs. It is also here where Balthasar van der Pol (1889-1959) invented the famous Van der Pol equations. Many years later, it was obvious to Gerton to send Philips an open application.

After four months he received an invitation for an interview. The interview did not take long. "They let me talk with five people. The funny thing was that they hardly asked me anything! They just told me what they were working on. I decided to simply listen and ask a question every now and then to show that I was interested. At the end of the interview they gave me the job."

Gerton goes on explaining that, interestingly, at Philips Research, you effectively sign your contract without knowing what exactly you will be working on. Namely, once Philips Research offers you a position, you still have to go through another round of interviews. In this second round you are told about specific project groups in detail. "I talked to people from audio signal processing, video signal processing, chip design, and modeling of transistor characteristics. At the end of this day, I could choose what interested me most and I decided to join the video signal processing group."

The Evoluon in Eindhoven

Working for Philips

Gerton's project had to do with so-called motion estimation. Ordinary televisions project 50 images per second on the screen. However, a 100Hz TV also receives only 50 images and the information must be doubled to 100 images by interpolation. A simple solution would be to simply show each image twice before moving on to the next. However, watching a person walk across the screen would then result in what is called judder. Namely, your eyes (unnoticably) adapt to the apparent motion in the images provided and start anticipating the position of the moving person on the subsequent images. This results in the motion being perceived as jerky, i.e. judder.

More advanced techniques use the information obtained from two subsequent images and interpolate between two positions. The best solution currently available uses linear interpolation between two images, but based on a low-resolution motion vector field on 8x8 pixel blocks. Hence, the vector field does not quite fit around the object, and especially with fast-moving objects, this results in a blurry region around the edges, which is called the halo effect. Can one do better, even with the restriction that, of course, this interpolation must be done in real time? There are also economic and manufacturing restrictions: the production costs of the chip must not exceed $10.

This is exactly what Gerton found appealing about the problem. During his mathematics studies he obtained a lot of skills and it was fun to apply these tools to a practical problem while dealing with all these extra non-mathematical restrictions. The intruiging part is that there is typically already a solution at hand. "These engineers have a really good intuition. They immediately see what the potential limitations are, both in terms of the technical and economic restrictions, and are very quick in proposing a feasible algorithm. However, this may not be a very good solution and certainly not the optimal one. A thorough more fundamental analysis, or even an investigation of how the problem behaves, is not their approach to such a problem. That was my work and I used it to try make the solution better, or propose a radical change to the algorithm."

Contrasts and similarities between University and Philips

Gerton has a tremendous respect and admiration for the organization within Philips Research. "Their way of working is totally different from what I was used to at the University of Groningen." First of all, the projects are a lot shorter, only six to twelve months, with long projects lasting no more than two years. "For each project, even those involving fundamental research, there are very clear agreements about the objectives, the time-frame, and assessment of intermediate goals. It is also clear who is responsible for the project and who assesses it." PhD research at a university can also be on a specific project with pre-defined objectives. However, according to Gerton, if the research does not seem to go anywhere, then this is the problem of the PhD student who should struggle on to get some results. If a project is not going well at Philips Research, then the group leader (who oversees about 45 people) can decide to put you to work somewhere else. "This man has an incredible skill to judge whether you are suitable for the project or not, and whether you would enjoy something else better. Obviously, this makes both you and Philips happy."

Gerton found the work and research at Philips in many ways very similar to PhD research at a university. "We were certainly expected to publish in international journals. There were group meetings and seminars, we had a very good library, and we were encouraged to go to conferences. If you had attended a conference, you were to report your experiences at the next group meeting. I thought this was a great way to inform others of new developments and it certainly made me attend conferences in a very different, much more active way. I would go to talks which I thought would be of interest not to me but to a colleague who could not attend!"

Back to academia

Despite the fact that working for Philips really was an extremely pleasant and motivating experience, Gerton decided after two years to move back to academia. "I do believe Philips is a great place to work. You know, there were people working for Philips whose fathers and grandfathers had also worked for Philps. However, I started to lose interest in the idea of just making better TVs." Even though, at the university, Gerton lost interest because he did not want to do research purely because of the beauty of mathematics, at Philips Research he realized that working towards a goal with specific applications was not motivating and challenging him enough. Gerton decided to apply for a postdoctoral position in Bioinformatics at Oxford University.

"Bioinformatics is about nature itself, about how evolution works. The task is to understand, extract and be able to put to use information from the human genome. The tools are... Everything you can think of as useful: statistics, stochastics, algebra, algorithmics, and a genuine appetite for solving puzzles." Gerton certainly wants to change the world and he enthusiastically begins to tell me about the human genome project. "Just imagine how nature did all that! This is research into the evolution of humans, of life itself." Furthermore, as he points out, the founder of the field, Mike Watermann, also started his research in dynamical systems. "I met him at Eurandom in Eindhoven while still working for Philips. Talking to him gave me the confidence that with my background, I could do something in this exciting field."