백석예술대학교 도서관

내용보기

Uncovering Higher-Order Structures in Complex Systems with Multivariate Information Theory- [electronic resource]

서명/저자: Uncovering Higher-Order Structures in Complex Systems with Multivariate Information Theory - [electronic resource]

주기사항: Source: Dissertations Abstracts International, Volume: 84-12, Section: B.

초록/해제: 요약Complex systems are defined by the presence of intricate, emergent structures that integrate many disparate elements into a single "whole." A central challenge of modern science is inferring this structure from limited and often noisy data. This thesis explores how information theory can reveal structured dependencies in data, with a particular focus on synergistic interactions: when there is information in the joint state of multiple variables (the "whole") that is inaccessible when considering the "parts" individually. Here, we explore three different mathematical approaches to assessing synergy in complex systems and what they reveal about the structure and dynamics of on-going brain activity at multiple scales. We find that higher-order synergies are widespread in the nervous system, existing at the level of local circuits of spiking neurons, as well at the level of whole regions of cortex. Synergistic information can also exist in the instantaneous functional coupling of elements, as well in higher-order flows of information through time. We find that existing methodologies from complex systems science are often insensitive to these synergies. We end by proposing the existence of a "shadow structure": a combinatorially vast space of dependencies that have gone unexplored due to the limitations of standard statistics.