#EA-Colloquium: Prof. George Kampis, German Research Center for Artificial Intelligence (DFKI), "Agent Based Simulations in Social and Biological Systems: Case Studies", 9 May 2017, 2.00 pm
Wednesday, 03 May 2017
In his lecture Prof. George Kampis of the German Research Center for Artificial Intelligence (DFKI) will review his work done using agent-based modeling (ABM) techniques in the last few years. He will start with models of ad-hoc communication (e.g. mobile phone communication without a signal), and discuss various trade-offs related to the choice of communication modes. The basic model is general enough to take any city map and to calculate behavior using assumptions about persons’ motion. A calibrated version uses real-world motion data from Zurich, collected during a city festival. He will briefly discuss anonymization by simulations using the statistical properties of sensitive personal data. As a second example then, he will discuss different variants of "collaborative classification" where agents (persons) interactively and incrementally define an emergent global classification based on their individual experiences. The problem setting is an abstraction again covering a number of cases from movie and restaurant rankings to smart city problems related to traffic and other aspects of city life. Classifier fusion is a well-known example in Artificial Intelligence and the dynamic incremental classifications in his study signify a new way of looking at this old problem. Finally, he will turn to a biological example where predator-prey interactions produce counter-intuitive results. Connected fragmented habitats turn out to facilitate species coexistence rather than degeneration - an unexpected upshot form a study that has since been experimentally justified. Prof. Kampis will finish by briefly mentioning his own works in data crawling, data analysis, and big data domains.
The colloquium will be held in English in our EA Lab from 2 to 4 p.m.