Spiking Neurons, Competing Agent Populations, and
Network QoS Through Distributed Control
Erol Gelenbe
Imperial College, London
Since the earliest days of computing, biologically inspired
paradigms have often served to develop computational models and
systems based on those models. Kleene's early paper on events in
nerve nets yielded one of the first mathematical studies of regular
expressions or finite state languages which are at the front end of
compiling systems, while John von Neumann was writing about the
synthesis of reliable organisms from unrealiable components which
provided the seed for research on hardware and system fault
tolerance. In this lecture we will discuss our own work on the
development two mathematical models which are inter-related: one of
them deals with interacting populations which have mutually
unpleasant behaviours as well as mutation, and the other model
describes spiked random neural networks. We will show how the
latter model has helped us design a "cognitive packet network" where
packets are routed based on users' QoS requirements. We will also
indicate how the model of interacting populations may help us
understand how to defend ourselves from Internet based attacks.
Wednesday Seminars
Click here for previous seminars.
|