Annual reports

Interacting stochastic systems consist of a large number of interacting random components. These components interact with each other and with their environment. Even when the interaction is local, such systems typically exhibit a complex global behavior, with a long-range dependence resulting in anomalous fluctuations and phase transitions. To mathematically understand these systems requires the use of powerful probabilistic ideas and techniques. The challenge is to introduce simple models, which serve as paradigms, and to unravel the complex random spatial structures arising in these models. Statistical physics and ergodic theory provide the conceptual ideas, while probability theory provides the mathematical language and framework. The important challenge is to give a precise mathematical treatment of the physics that arises from the underlying complexity. Much of the knowledge that has been built up in mathematical statistical physics over the past decades is currently making its way into biology. One of the tasks is to help facilitate this cross-fertilization and to address concrete biological questions at the interface. Examples are coming from population genetics and immune system biology.

The research in Probability Theory concentrated on interacting stochastic systems (disordered systems, percolation, random polymers, metastability, sandpiles), ergodic properties of random processes (dynamical Gibbs-non-Gibbs transitions, hidden Markov chains), and topics from mathematical biology (population dynamics, T-cells). Key tools are large deviation theory, stochastic analysis, variational calculus and combinatorics.  There is an interesting link between algebraic dynamical systems and solvable models of statistical mechanics. It turns out that entropies of apparently different systems often coincide, and that this `mere' coincidence is not accidental. Research aims at providing an explanation for this phenomenon. A powerful combinatorial technique to study high-dimensional systems is the 'lace expansion'. Research aims at obtaining a rigorous understanding of phase transitions in high dimensions, including diffusion on critical spatial structures.

The research in Operations Research concentrated on Markov chains, Markov decision processes and Markov games, with applications to problems in stochastic networks. One of the main issues concerns stability. How can stability be checked? If stable, how fast does the network reach its stationary distribution? If unstable, what does the quasi-stationary distribution look like? How can efficient algorithms be developed to control the network according to certain pre-set optimization criteria? Are these algorithms amenable to practical implementation? What can one say about the structure of optimal policies? Which type of customer should be prioritised to optimise network performance? These questions can be studied within the framework of Markov chain theory.

Often the situation arises where there are conflicting interests, for instance, maximizing server efficiency while minimizing customer dissatisfaction. This may be studied through Markov game models.