Causality from phases of high-dimensional nonlinear systems

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F67985807%3A_____%2F25%3A00604108" target="_blank" >RIV/67985807:_____/25:00604108 - isvavai.cz</a>

Výsledek na webu

<a href="https://doi.org/10.1016/j.ins.2024.121761" target="_blank" >https://doi.org/10.1016/j.ins.2024.121761</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1016/j.ins.2024.121761" target="_blank" >10.1016/j.ins.2024.121761</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Causality from phases of high-dimensional nonlinear systems

Popis výsledku v původním jazyce

Detecting causal relations in large dynamical systems is a difficult endeavor. As system dimension increases only linear approaches can be used, under the not always reasonable assumption of linear dynamics, since applications of nonlinear approaches are computationally intractable. Herein we test our recently developed information theory-based approach for causality detection from phases, appropriate for time series that exhibit well-behaved oscillatory patterns, and we expand our previous analysis to large systems with intricate connection patterns. Other than periodic-like behavior, this approach does not require any assumptions and works well for large-dimensional systems. We assess its performance on artificial data from networks of 3 or 10 coupled Rössler oscillators and networks of 3 coupled Mackey-Glass equations. We then employ it to study the dynamics of the human brain in two test-cases, one of emotional state change in healthy subjects and one of pathological system change in epilepsy. In the course of our study, we identify some very interesting phenomena related to synchronization, which can lead any causality measure to failure. We finally discuss the steps needed to properly investigate causal relations, so that even if some real connections cannot be detected, at least false connections would not be inferred.

Název v anglickém jazyce

Causality from phases of high-dimensional nonlinear systems

Popis výsledku anglicky

Detecting causal relations in large dynamical systems is a difficult endeavor. As system dimension increases only linear approaches can be used, under the not always reasonable assumption of linear dynamics, since applications of nonlinear approaches are computationally intractable. Herein we test our recently developed information theory-based approach for causality detection from phases, appropriate for time series that exhibit well-behaved oscillatory patterns, and we expand our previous analysis to large systems with intricate connection patterns. Other than periodic-like behavior, this approach does not require any assumptions and works well for large-dimensional systems. We assess its performance on artificial data from networks of 3 or 10 coupled Rössler oscillators and networks of 3 coupled Mackey-Glass equations. We then employ it to study the dynamics of the human brain in two test-cases, one of emotional state change in healthy subjects and one of pathological system change in epilepsy. In the course of our study, we identify some very interesting phenomena related to synchronization, which can lead any causality measure to failure. We finally discuss the steps needed to properly investigate causal relations, so that even if some real connections cannot be detected, at least false connections would not be inferred.

Druh

J_imp - Článek v periodiku v databázi Web of Science

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

<a href="/cs/project/GF21-14727K" target="_blank" >GF21-14727K: Struktury synchronizace v mnohorozměrných neurálních signálech: strojové učení a predikce účinnosti antidepresiv</a><br>

Návaznosti

I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Rok uplatnění

2025

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název periodika

Information Sciences

ISSN

0020-0255

e-ISSN

1872-6291

Svazek periodika

697

Číslo periodika v rámci svazku

April 2025

Stát vydavatele periodika

NL - Nizozemsko

Počet stran výsledku

19

Strana od-do

121761

Kód UT WoS článku

001409475600001

EID výsledku v databázi Scopus

2-s2.0-85212337184