student profile: Leonardo Novelli


Map

Thesis work

Thesis title: Inferring information network structure from dynamics

Supervisors: Mikhail PROKOPENKO , Joseph LIZIER

Thesis abstract:

The increasing availability of large-scale and fine-grained datasets provides an unprecedented opportunity for the advancement of quantitative and computational studies of complex systems. Given the massive scale of the available data sets, a shift towards a data-driven modelling of complex systems requires efficient algorithms for analysing multivariate time series, obtained from observation of the activity of a large number of elements.

My thesis will focus on model-free algorithms for network inference, which can be applied to multivariate time series and aim at constructing the simplest directed and time-dependent network model which can reproduce the activity of a system. This approach is particularly suited for the study of biological and socio-economical systems, whenever the time series are available but a model of the interactions needs to be built.

The field of information theory provides powerful tools and measures for this effort. In particular, transfer entropy has been recognised by the research community as a natural fit for effective connectivity inference, since it provides a directional, dynamic, model-free measure of dependence of a source on a target variable.

Standard approaches based on transfer entropy use statistical significance to determine whether links should exist. Several enhancements have been recently proposed; in particular, iterative or greedy approaches with conditional transfer entropy can both capture synergies and eliminate non-required redundancies, while avoiding combinatorial explosions in the number of source combinations considered. There are subtle differences between these techniques, and many of their features remain unclear. At the start of my project, I will explore their strengths and weakness, aiming to establish their range of applicability. I will also seek to enhance these algorithms by incorporating findings on the relationship between network structure and dynamics that will be established in the wider ARC DECRA fellowship project “Relating function of complex networks to structure using information theory”.

Selected publications

Download citations: PDF RTF Endnote

Conferences

  • Wollstadt, P., Lizier, J., Vicente, R., Finn, C., Martinez Zarzeula, M., Lindner, M., Martinez Mediano, P., Novelli, L., Wibral, M. (2017). IDTxl - The Information Dynamics Toolkit xl: a Python package for the efficient analysis of multivariate information dynamics in networks. Bernstein Conference 2017, Germany: Bernstein Conference. [More Information]

2017

  • Wollstadt, P., Lizier, J., Vicente, R., Finn, C., Martinez Zarzeula, M., Lindner, M., Martinez Mediano, P., Novelli, L., Wibral, M. (2017). IDTxl - The Information Dynamics Toolkit xl: a Python package for the efficient analysis of multivariate information dynamics in networks. Bernstein Conference 2017, Germany: Bernstein Conference. [More Information]

Note: This profile is for a student at the University of Sydney. Views presented here are not necessarily those of the University.