Collective Behaviour - Summer Seminar Series 2021

Reservoir computing for swarming individuals and swarming groups


Shannon Algar, Forrest Research Foundation

Shannon Algar is a Fellow at the Forrest Research Foundation in Australia, where she works using applied mathematics to understand complex systems. Her current work will use swarm intelligence to analyse the health and fitness of animals with the aim of improving their welfare.

Reservoir computing for swarming individuals and swarming groups

In this talk I will introduce the use of reservoir computers in combination with agent-based models of swarms as our group has used them in two different scenarios: 

Firstly, to demonstrate that a previously proposed movement rule for the Selfish Herd was reasonable in a biological setting, we show that it can be learned by individual agents using the most basic of neural networks. This suggests that a more sophisticated neural network, such as a brain, could also learn this computationally complex (but intuitively simple) behaviour. 

Secondly, we outline a framework that uses the entire swarm as a reservoir, effectively illustrating how a flock of birds could be used as a computer. The relationship between the flock’s performance as a reservoir computer and the behaviour of the group demonstrates that optimal computational properties are obtained near a phase transition regime.

Datum: 2021-05-10