Mehr lesen
This book introduces a unified framework that integrates various data-driven information dynamics approaches to quantify node-specific, pairwise, and high-order interactions within complex systems in the contexts of network neuroscience and network physiology. Using measures of information rate, a hierarchical organization of interactions is established to describe the dynamics of individual nodes, connections between pairs, and redundant or synergistic relationships among groups of nodes. Initially defined in the time domain, these measures are extended to the spectral domain, enabling frequency-specific analysis under the Gaussian assumption and linear parametric models. The framework is validated on simulated network systems and applied to real-world multivariate time series in neuroscience and physiology. The spectral high-order information measures successfully reveal respiratory-driven redundancy in cardiovascular, cardiorespiratory, and cerebrovascular systems, and uncover a predominance of redundancy in high-order brain interactions, alongside the emergence of synergistic circuits not captured by pairwise analysis. These results emphasize the importance of high-order, frequency-resolved information measures in characterizing complex network dynamics and provide new insights into the coordinated functioning of physiological and neural systems.
Inhaltsverzeichnis
Introduction.- Linear Modelling of Stochastic Interactions.- Static Networks of Random Variables.- Dynamic Networks of Random Processes.- Applications to Physiological Networks.- Applications to Brain Networks.
Zusammenfassung
This book introduces a unified framework that integrates various data-driven information dynamics approaches to quantify node-specific, pairwise, and high-order interactions within complex systems in the contexts of network neuroscience and network physiology. Using measures of information rate, a hierarchical organization of interactions is established to describe the dynamics of individual nodes, connections between pairs, and redundant or synergistic relationships among groups of nodes. Initially defined in the time domain, these measures are extended to the spectral domain, enabling frequency-specific analysis under the Gaussian assumption and linear parametric models. The framework is validated on simulated network systems and applied to real-world multivariate time series in neuroscience and physiology. The spectral high-order information measures successfully reveal respiratory-driven redundancy in cardiovascular, cardiorespiratory, and cerebrovascular systems, and uncover a predominance of redundancy in high-order brain interactions, alongside the emergence of synergistic circuits not captured by pairwise analysis. These results emphasize the importance of high-order, frequency-resolved information measures in characterizing complex network dynamics and provide new insights into the coordinated functioning of physiological and neural systems.
