Preface
The idea emerged of organizing a PPSN conference at the University of Dortmund when a group of us including Heinz Mühlenbein and others became aware of an opportunity. Occasionally, but more and more often over the years, we met each other on the outskirts of classical conferences. Societies of mathematicians, physicists, computer scientists, or systems analysts and operations researchers had added special sessions covering ''new paradigms'' like synergetics, adaptive learning, and complex phenomena, to their annual meetings. But we did not feel we belonged to the kernel of these subpopulations, and we began to recognize that we ourselves, though coming from different scientific disciplines, formed a new subpopulation-a classical case of annidation within ecosystems.
We decided to enhance our field of common interest by gathering people with similar research goals regardless of how they were embedded in special disciplines. However, it required some effort to formulate that common interest and to coin an unbiased term like ''Naturanaloge parallele Problemlösungs-Strategien'' (in German), which David Goldberg helped to translate into ''Parallel Problem Solving from Nature (PPSN)''.
With the appearance of massively parallel computers, increased attention has been paid to algorithms which rely upon analogies to natural processes. This development defines the scope of the PPSN conference at Dortmund in 1990. The scope included but was not limited to the following topics:
The scientific programme committee consisted of:
The committee helped to select for presentation what were hopefully the most interesting among about 100 proposals. To avoid excessively strong selection-which is a strategic prerequisite when solving hard optimization problems-we changed our minds and switched from 30 full length oral presentations to just 22, with reduced time for the speakers, plus 44 poster presentations. This change caused only very few colleagues to withdraw their proposals.
During the three days of the workshop and especially during the closing session there was a broad consensus for gathering again. Taking into account the partly related topics of the well established International Conference on Genetic Algorithms, which has been held in the U.S.A. every odd year since 1985, many participants thought PPSN should be held every even year somewhere on this side of the Atlantic Ocean. In view of the overwhelming response to this first PPSN we are looking forward to holding a perhaps even larger PPSN II conference during the year 1992.
It is a pleasure for us to express our thanks to those organizations who acted as sponsors of PPSN I: Parsytec GmbH and Paracom GmbH, Aachen Society of the Friends of the University of Dortmund e.V., Dortmund IBM Deutschland GmbH, Stuttgart Siemens AG, Munich The Computer Society of IEEE Ministry of Science and Research of Nordrhein-Westfalen, Düsseldorf All their funding was used to help pay the travelling expenses of those active participants who would otherwise have been unable to come.
We would also like to thank all those who helped in preparing and implementing the conference, and in completing it by organizing these proceedings. We would like to mention especially the co-chairman of the programme commitee, Prof. David E. Goldberg, as well as all the members of the local organizing committee:
We also thank Antje Schwefel as well as other collaborators at the Chair of Systems Analysis of the University of Dortmund, without whose enthusiastic engagement the whole event could not have taken place. Last but not least, we give grateful thanks to our banquet speaker, Prof. Peter Bock from The George Washington University, who entertained us with his very lucid and bold perspective on the possible future of intelligent machines.
Heidelberg Reinhard Männer
Dortmund Hans-Paul Schwefel
February 1991
As long as computers could not perform more than one simple instruction and handle more than one or two items at a time, it seemed to be crazy to imitate natural multi-particle, multi-cellular, or even population processes as programming paradigms in order to solve problems. Those who came up with such ideas in the 1960s were laughed at. Yet there were a few who did what people have often done at the beginning of an era of a new technology, namely look for prototypes in nature. Hans J. Bremermann's ''search by evolution'' [1], John Holland's ''genetic algorithm'' [2], and Ingo Rechenberg's ''Evolutionsstrategie'' [3], to mention only a few, where apparently forgotten or neglected when the first MIMD (multiple instructions, multiple data) computers became available to a larger public. Sometimes it happens that history later centers on positions which were outlying before, and if those who attended to the outlying positions have not meanwhile died or changed their mind, they may enjoy the reward of late acceptance.
It is a matter of fact that in Europe evolution strategies and in the U.S.A. genetic algorithms have survived more than a decade of non-acceptance or neglect. It is also true, however, that so far both strata of ideas have evolved in geographical isolation and thus not led to recombined offspring. Now it is time for a new generation of algorithms which make use of the rich gene pool of ideas on both sides of the Atlantic, and make use too of the favorable environment showing up in the form of massively parallel processor systems, such as transputer farms, large LANs (Local Area Networks) and other new computer architectures. There are not many paradigms so far to help us make good use of the new situation. We have become too familiar with monocausal thinking and linear programming, and there is now a lack of good ideas for using massively parallel computing facilities.
We have so far emphasized only the imitation of evolutionary population processes , but this does not mean that other natural metaphors are not worth considering. On the contrary, the effectiveness and the normal efficiency of the human immune system is a further example of a collective process which is worth investigating if we look toward solving adaptation instead of optimization tasks. Neural nets as models of brain processes should be mentioned at this point, since they have been accepted as paradigms for naturally intelligent processes from the very beginning of the computer era. ''Design for a Brain'' [4], an early book of Ross Ashby, and ''The Computer and the Brain'' [5], an early book of John von Neumann, testify to the human hubris that we like to believe more in the power of individual than of collective intelligence.
There may be a link, unfortunately missing during this conference, between both approaches. Michael Conrad [6] and Gerald Edelman [7] have given hints, at least, that neural processes may also be interpreted in terms of an evolutionary agenda. A strong argument for neglecting neither collective against individual intelligence nor explorative search against logic reasoning, which is always based upon limited knowledge only, should be the general observation that human beings are a product of evolution and not its motor, and insofar as they have acted as a motor this has driven our planet into a dangerous situation.
The chapters of these proceedings have been arranged in a way which completely neglects the sequence of the conference lectures. We are very grateful to Thomas Bäck who has looked for similarities and differences within the altogether 65 contributions and thus found a simple but hopefully helpful hierarchical arrangement. A final statement of thanks goes to the authors who have patiently submitted to the process of halfway streamlining the layout of this volume, as well as to all those who contributed to the lively discussions during the conference and thus challenged further investigations, the outcome of which we hope to see and hear at PPSN II.
References