The cluster is funded by the EPSRC "Novel computation, coping with complexity" initiative. The general themes of the cluster are the search for overarching principles which apply in complex natural systems such as geophysical and ecolgical systems, and how such principles might be exploited for novel computation.
Many computing approaches are based on agents or objects - NANIA has adopted the term autonome to refer to any interacting multi-state system which can encompass cellular automata, agent, organisms or species.
Open Questions in Complex Systems
The scientific theme of this proposal is the search for global principles with which to describe or control complex autonome systems. This extremely ambitious goal is likely to be achieved only in part: such open systems can be subject to arbitrary external influences, and there are simply too many types of open complex system to expect a unique paradigm. Generalised principles will emerge for sub-classes of systems from studying rigorously defined classes of simple systems computationally. These systems will be motivated by and compared with real systems, and consideration of them will serve to illuminate and assist the more specific science of the workplans.
The understanding sought in NANIA of both where a complex system is likely to fail stressed links/autonomes, and where such failure would be most destructive key links/autonomes - this could have tremendous benefit in systems design. Distributed computing presents problems of both security and robustness - a better understanding of this could enable computer or network design which could tolerate large hardware failure rates through most of the system with only a few nodes crucial to operation, and therefore demanding the highest levels of manufacturing quality and maintenance. The current view of these is that they are the highly connected nodes of a network, but in ecosystems (and earthquakes) the key species (failure regions) have no such obvious properties, and may shift in time.
In particular, we seek to address the following overarching scientific questions:
- When can we define a global quantity that a given autonome system optimises?
- How can the information-theoretic entropy be applied to complex systems where the bit-equivalent is ill-defined?
- When does a system produce a diverse population rather than a replicated "optimal" autonome and what does this imply about cooperative behaviour?
- What controls the flow of resources through a system?
- How much complexity is required for a system to convert steady driving into scale-free release?
- When and how can we coarse-grain models without losing the essential nonlinear complexity?