Research Supervisor Connect
The research will involve computational analysis in complex systems, complex networks, information theory and collective animal behaviour. The student will be exploring new algorithms for measuring multivariate information flow relationships between animals in flocks/schools, based on time-series recordings of their positions and speed. The PhD will be supervised by A/Prof. Joseph Lizier. The applicant will join A/Prof. Lizier’s Information Dynamics team in the Modelling and Simulation group, which studies complex systems and networks at The School of Computer Science. The student will collaborate with Prof. Ashley Ward (Life Sciences), within the University’s Centre for Complex Systems.
Associate Professor Joseph Lizier.
Collective behaviour such as flocking or schooling in animals such as birds and fish provides important evolutionary advantages, such as predator avoidance. Understanding the dynamics of these group behaviours are critical to inform our management of a changing environment. Fully characterising the connections between individuals in these groups remains difficult though, since flocking or schooling interactions induce fluid structures whereby individuals who are close-by and genuinely interacting may shift over extremely short periods of time. This presents substantial challenges for standard time-series analysis, yet if we could wholistically model such interactions there is major interest in the utility of such a method.
As a first step, we have adapted the information-theoretic measure transfer entropy (TE) to provide the gold-standard approach for measuring pairwise information flows in flocks/schools. This method identifies source-target pairs which are pairwise interacting at any snapshot in time In collaboration with Prof. Ashley Ward (USyd Life Sciences) we have used this method to relate pairwise TE to hunger and predation in fish schools, and relative location of source fish [24]. This project will extend that research beyond pairwise information flows to measure higher-order interactions in flocks/schools (e.g. to detect when an effect on a target results from a synergistic combination of two source individual’s actions). This will allow us to infer for the first time the set of causal parents for a target individual at any given snapshot in time. The project will utilise our JIDT (https://github.com/jlizier/jidt) open-source toolkit, extending the methodology for studying information flows in Flocking within it. There is potential for many applications to real data sets of schooling fish.
Successful candidates must have:
How to Apply:
To apply, please email to joseph.lizier@sydney.edu.au the following:
A cover letter (or paragraphs in the email) explaining your interest in, and suitability of skills/background/experience for, this project. Please highlight your academic results, any published papers and research/industry experience.
The opportunity ID for this research opportunity is 3423