A Theory of Agent-Based Models

12 settembre 2017
Versione stampabile

Venue: Seminar Room “-1” – Department of Mathematics - Via Sommarive 14 - Povo -Trento
at 11:00 a.m.

  • James F. Lynch (Clarkson University)

Abstract: 
A philosophical question of considerable practical importance is what distinguishes biological systems from man-made computational systems. This question occupied John von Neumann in the last years of his life, and led to his invention of cellular automata. Alan Turing was also interested in this question, and he studied pattern formation in organisms. His ideas have recently been corroborated by the discovery of the biochemical networks that control pattern formation. More recently, computer scientists, including Cardelli and Zavattaro, have addressed this question. All of the models used by these researchers fit within the paradigm of agent-based models (ABMs).

We will briefly review these ideas, and then give a mathematical definition of ABMs. We propose that ABMs are stochastic processes whose states are metafinite models. Since metafinite models are not well-known outside a small community of computer scientists and logicians, we will describe them in a way that we hope anyone with some experience in graphs will follow. We give some examples from systems biology, and describe some algorithms for simplifying ABMs without losing important information about their behavior.

Contact person: Ozan Kahramanogullari