April 23rd, 2019, Faculty of Computer Science and Electrical Engineering, University of Rostock
Typically in answering a research question via simulation studies, not one but different simulation models are built. This may happen as a by-product of iterative calibration, validation, and model revising phases. Here, the intermediate products are often considered of little relevance, although they document where one model fails and another succeeds in explaining observations. In some cases, it is clear from the outset that multiple models will be developed and thus are deliberate products of the simulation study. This happens, for example, when different models are coupled in multi-scale or multi-resolution approaches and form model families to be adapted to concrete data, or constitute model reductions to handle the complexity.
The one day workshop will start with a talk by Bernard P. Zeigler (as part of the IEEE Computer Society Distinguished Visitors Program, https://www.computer.org/web/chapters/dvp)
WHY SHOULD WE DEVELOP SIMULATION MODELS IN PAIRS?
Bernard P. Zeigler
Professor Emeritus,
Department of Electrical and Computer Engineering
University of Arizona
and
Chief Scientist,
RTSync Corp.
ABSTRACT: The conventional approach to model construction for simulation is to focus on a single model and follow a more or less structured development cycle. Why should we put in twice the time and effort to develop two models rather than one? The answer lies in the fact that like most greedy heuristics, short-sightedness at the beginning may be much more costly in the end. This talk champions the cause of the pairs-of-models approach. We show how this approach eventually leads to better results than initially attempting to construct a complex model, followed later by having to revert to a simpler model when increasing complexity makes progress hard to achieve. We show how pairs-of-models development can be supported by computational tools for relating structure and behavior between models. Benefits of pairs of models, and eventually families of models, include the ability to perform mutual cross-calibration, avoid the usual difficulties in harmonizing the underlying ontologies, and narrow the search for plausible parameter assignments.
Thereafter two sessions are planned with short experience talks from diverse application fields such as demography, electrical engineering, and cell biology. The workshop will conclude with a panel session to discuss methodological challenges arising from developing and managing multiple simulation models.
Bernard Zeigler will leave on Tuesday late afternoon to visit the Applied University of Wismar where a follow up program will be organized on Wednesday morning.
If you have further questions, please contact adelinde.uhrmacheruni-rostockde.
If you would like to contribute an experience talk, please send title and abstract until March, 15st 2019 to adelinde.uhrmacheruni-rostockde.
If you would like to participate in the event, please send an e-mail to Sigrun Hoffmann sigrun.hoffmannuni-rostockde until April, the 1st 2019.
A more detailed program and the room (Campus South, University of Rostock) where the event will take place will be published March, the 25th 2019.