Report

Teaser, summary, work performed and final results

Periodic Reporting for period 2 - COSMOS (Complex Oscillatory Systems: Modeling and Analysis)

Teaser

Complex oscillatory systems are abundant in nature, physical and engineering devices, and life sciences. The overarching aim of COSMOS is to understand the behavior of such large and complex systems, specifically those composed of multiple interrelated subunits, many of which...

Summary

Complex oscillatory systems are abundant in nature, physical and engineering devices, and life sciences. The overarching aim of COSMOS is to understand the behavior of such large and complex systems, specifically those composed of multiple interrelated subunits, many of which operate on different time scales. The novel interdisciplinary approach of COSMOS is to combine theoretical techniques with data-analysis procedures, to allow the development and validation of original analysis methods for complex systems.
COSMOS is pursuing three main objectives:
• The development of methods for the identification of the relevant variables describing collective dynamics of coupled oscillators (top-down)
• The extraction of relevant information from multivariate-data recorded in complex oscillatory systems (bottom-up)
• Integration of the two approaches for cross validation and refinement; implementation of the findings in software toolboxes for the use of unskilled users

Work performed

The activities of the ITN begun with a school and workshop in Florence (November 2015) focused on bridging previous background to the topics of COSMOS. The second school in Aberdeen (June-July 2016) aimed at advance topics of theory of complex systems and on transferable skills. At the workshop in Amsterdam (December 2016) and at the retreat in Wittenberg (March 2017) ESRs reported on their first results and had opportunity to discuss topics related to industrial applications with representatives of industrial partners of the consortium. The Toolbox Workshop (Brijuni, October 2017) was devoted to a practical exchange of the algorithms and methods
developed. The highlight of the common activities was an international conference “Analysis and Modeling of Complex Oscillatory Systems” (AMCOS) held at Barcelona (March 2018). It was completely prepared and conducted by ESRs. This conference attracted world-leading researchers (altogether 119 attendees from 20 countries), contributed enormously to visibility of the COSMOS research and provided invaluable experience in essential transferrable skills to all ESRs. Final conference (Novo Mesto, September 2018) was devoted to discussions of the projects and potential fields of further collaborations.
During the 2nd reporting period, the ESRs advanced enormously in their research projects, as documented in 39 publications in leading peer-review journals, and a total of 71 presentations at different Conferences. All ESRs participated at different types of outreach activities, ranging from talks for high-school students, to youtube tutorials.

Final results

The main goal of the “top-down” work package was to combine tools from nonlinear dynamics and statistical physics to describe complex oscillatory systems by virtue the dimension-reduction techniques. A substantial progress has been achieved in this direction, resulted in about 18 publications in high-quality journals by ESRs. A very hot topic was study of chimera states, which combine properties of synchronous and asynchronous regimes. Here, ESR10 established properties of phase and generalized synchronization between chimera states, while ESR1 discovered and described a novel state of blinking chimera. In studies of neural field networks, ESR7 developed a diffusion approximation for a stochastic model of neural activity, and ESR12 juxtaposed properties of scale-freeness and of partial synchrony for neural mass networks. Effects of noise on the dynamics of complex oscillatory networks were in focus of several projects. ESR7 studied desynchronization and pattern formation in a noisy feed-forward oscillatory network. He also proposed and studied a simple stochastic model of neuronal excitatory and inhibitory interactions on a directed lattice. ESR2 discovered a nontrivial mechanism of noise-induced stabilization of collective dynamics in oscillator populations. In a more general setup, studies by ESRs 5 and 3 showed that nonautonomous systems can have properties quite different from that of standard autonomous ones. An important theoretical advance in the analysis of complex oscillatory networks was made by ESR2, who created a novel dimension reduction technique that includes not only the phases, but also variations of the amplitudes. ESRs 12 and 13 established existence of first-order phase transitions in the Kuramoto model with compact bimodal frequency distributions. Two other studies dealt with effect of delay. ESR11 described the dynamics of a large system of spiking neurons with synaptic delay. ESR3 found evidence of a critical phase transition in a purely temporal dynamics with a long-delayed feedback.
In the frame of “bottom-up” work package, methods of inference have been developed and applied to synthetic and real data. One overlapping topic was characterization of spike trains. In a series of papers, ESR17 developed methods of reconstructing networks of pulse- coupling oscillators, without any preliminary knowledge about their properties. Work of ESR8 was concentrated on the characterization of similarities and diversities in spike trains. Furthermore, he developed measures of spike train synchrony for data with multiple time scales, and quantified consistency in spatio-temporal propagation patterns. ESR10 characterized robustness and versatility of a nonlinear interdependence method for directional coupling detection from spike trains. Another overlapping topic was network reconstruction. Work of ESR4 concentrated on the analysis of degree distributions of reconstructed networks in dependence on the statistics of false positive and false negative
conclusions about the links. ESR14 and 10 collaboratively studied, how a rank-based connectivity measure can be used to infer the network connections. Furthermore, ESR14 developed a method of evolutionary optimization of network reconstruction from derivative-variable correlations. A problem of cardio-respiratory interactions was attacked by ESR16. He developed a method of disentangling respiratory and non-respiratory components in the heart phase dynamics, and applied this to real physiological measurements. Significance is very important in the practical data analysis. Here ESR6 contributed by comparing different surrogate data techniques for hypothesis testing. In the frame of integration and comparison of different approaches, the consortium delivered 9 publicly available software packages summarizing created methods and techniques.
The ITN-EJD has contributed to train specialists in complex oscillatory systems, able to cope with different challenging setups fr

Website & more info

More info: http://www.cosmos-itn.eu/.