Samia Nefti-Meziani

University of Salford, UK

Non-rational Particle Swarm Optimization

Summary

Among Swarm Intelligence Systems, Particle Swarm Optimization was introduced in terms of social and cognitive behaviour. It is a promising new optimization technique which models the set of potential problem solutions as a swarm of particles moving about in a virtual search space.

However, in most PSO algorithms, there is only information sharing between particles, and recently a rational model has been introduced in each agent’s decision committed to maximising the accumulated total wealth. However, for a search which involves identification and location of the targets across a search area that is often vast and possibly dangerous, these particles should be able to take into account the uncertainty and the risk associated with a decision taken and also able to reproducing their preferences over a simple choice problem. Hence, these particles need to behave similarly like humans and be able to take effective decisions under risk-based conditions.

In my talk, I will show that endowing particles with some irrationality, unlike the already established paradigms of swarm intelligence, and using the prospect theory as a basis can improve significantly the search capabilities in risky and constrained environments.

Resume

Samia Nefti is currently Associate Professor at the School of Computing, Science and Engineering at the University of Salford, and the director of the Salford Advances Robotics Research Centre. She holds a Doctorate in Computer Science and DEA in Robotics from Paris XII, France. Her research interests over the last 19 years concern the development of cognitive models and intelligent information. Her contributions have focused on machine learning and the development of a novel clustering techniques for behaviours generation and understanding for robotics platforms. She is a Member of the IEEE, a Chartered member of BCS and an active member of the European Network for the Advancement of Artificial Cognitive Systems. She has been a program committee member of over 30 conferences and has acted as a program chair of several national and international workshops.