Early Computers and the Like

A career centered on improving the effectiveness of computers

 Irene Gargantini Strybosch, professor emeritus with Western University, IEEE senior member

The first contact I had with computers left me with an unforgettable impression. It happened at the Politecnico di Milano where I followed a course offered by, among others, Drs. Luigi Dadda and Lorenzo Lunelli. Could that machine with blinking diodes and triodes produced results? I joined the group who worked steadily on the CRC 102A,* an early computer by the Computer Research Corporation (later bought by NCR). The input was via a punched tape, it understood instructions written using numbers in base eight (000, 001, …,111); direct instructions included store, add, subtract and multiply, but not divide. It was a challenge to write short programs like the evaluation of a trigonometric function or finding the inverse of a 3×3 matrix.

The knowledge I acquired there was invaluable. It got me a job with Agip Nucleare first, followed by one with The European Atomic Energy Community (Euratom). Here I was involved in the calculation of the neutron flux, an essential task in this field, since a controlled amount of neutrons has to be produced at all times in order to keep a nuclear reaction stable. Numerical calculations played an essential role and these were done with computers of the IBM700-series (not using transistors yet). The importance of finding new methods that could speed up the massive computation while saving memory hit me as an essential task that should be pursued with consistency and vigor. At that time, neutron flux was evaluated by slices; the output (under the form of punched cards) of one layer became the input of the next one, each slice aligned along the vertical axis. One of the time-consuming computations was centered on the integral of some modified Bessel functions, known as Bickley functions. This raised an interesting question: could these functions be evaluated mathematically in forms other than the usual polynomials? I talked over with the group I was working with, but only one member seemed interesting in the issue. The first thing was to explore which approximations were available in the literature outside the polynomials: I encountered articles on Padé approximants, continued fractions and, in a recent paper by Fraser and Hart, a method resulting in very compact rational functions obtained by using the equalization of maxima, a procedure known as Remez algorithm. So T. Pomentale and I applied this novel approach to the functions essential to the evaluation of neutron flux. Once this done, one had to prove that the numerical values arising from the new approximations were correct. To this end I performed exhaustive (and exhausting) visual checks using the existing Mathematical Tables. Once convinced that our approach produced correct results in a fraction of the time used by methods on the market, I searched a journal interested in publishing our paper. One of the few journals available at that time was the Communications of the ACM. Our paper was accepted in record time.

As I advanced my career by joining The IBM Research Laboratory in Rueschlikon (Switzerland) I came across other interesting research topics; some concerned other forms of approximations of complex mathematical functions; another was to research constructive methods for the determination of polynomial zeros, as the lab was offering a symposium on the subject.

The next advancement in my scientific career was to accept a position with The University of Western Ontario, where I was able to conduct research amidst heavy teaching. The course overload was due to the fact that the university was offering the first undergraduate program in computer science (first in the country), which attracted a far more number of students than the department could handle. It was at that time that I investigated the iterative methods used to calculate polynomial roots. Was there a way to improve the convergence, i.e. making the evaluation faster? Computer time was still at a premium in the early ’70. This resulted in a seminal paper (Circular Arithmetic for the Solution of Polynomial Equations) that was the foundation of the several hundred articles that followed. I introduced a very original idea for iterative methods: given an approximate value, iterate not only on the starting point(s), but also on the error bound(s). In several occurrences, like in Newton’s and Laguerre’s, this approach improved the order of convergence by one unit (from quadratic to cubic, etc). This paper was the result of my own thinking, in spite of the coauthorship.

When the ’80 came, other exciting fields emerged; one of these was the demand of representing and searching spatial data effectively. My paper, entitled An Effective Way to Represent Quadtrees won the cover of the Communications of the ACM. This technique was used in structuring geographical data and, later on, applied effectively to the display of images—something extremely useful, since computer graphics was coming of age. While teaching this topic, I was approached by Dr. Uldis Bite, a medical doctor extremely interested in adapting some of my findings to medical imaging. He supplied me and my graduate students with several digitized tomography images; in 1988 one of the first software packages, “CTpak: An Interactive Utility for CT Data” was produced as part of the Master’s thesis of one of my students. It became routinely used at the Robarts Research Institute. Uldis was very busy, with his work and because of an incurable disease that hit his little girl; meanwhile I became chair of the computer science department and thus we never got around to publish together a scholarly paper on the subject. When Uldis left for a research post at the Mayo Clinic, I didn’t find a replacement of his caliber—an essential element to continue to be at the frontiers of this field.

Mandatory retirement was on the horizon; I tried to see if my expertise could be useful in other activities, like spreading the gospel about how important science and engineering are to the wellbeing of our daily life; my attempts didn’t find any interest within the university.

Meanwhile my husband’s health started to deteriorate; he was very happy I stayed home.

I found other interests—but this is another chapter of my life…

 

*www.dbpedia.org/page/Luigi_Dadda

Advertisements