Aristotle University Of Thessaloniki (AUTH)
Συνδεσιμότητα και δίκτυα χρονοσειρών με εφαρμογή στην επίδραση του διακρανιακού μαγνητικού ερεθισμού
Abstract
dc:descriptionThe study object of this thesis is the evaluation and development of new methods regarding the study of the interdependence structure of systems located behind a multi-dimensional time series. The concept of Granger causality has been used widely for the estimation of causality relations between the observed variables of a system. In this study a development of measures has been realized for the estimation of Granger causality based on dimension reduction. The new indices are differentiated: a) as of their applying domain (time domain, frequency domain), b) as of their base (models, correlation) and c) as of the kind of the causality relations that they track (linear, non-linear). After an extensive bibliographic review, the new measures are compared with a big number of causality measures that are widely used in the research of multi-dimensional time series. For the evaluation of the causality measures dynamic systems are used that came from stochastic processes, delay differential and non-linear difference equations. The connectivity structure of the systems can be represented as a network of which the connections are defined from the system equations. Therefore, knowing the true causal relationships, the performance indices are calculated for the evaluation of the causality measures. The main conclusion of this study is that the measures that are based on the dimension reduction are superior regarding the estimation of the connectivity structure. Finally, the measures were applied in the field of neuroscience and specifically in electroencephalographic recordings of patients with epilepsy that contain epileptic discharges that are terminated after using transcranial magnetic stimulation. The changes of the brain connectivity before, during and after the epileptic discharge are clear with the measures that are based in dimension reduction.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2021
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Singiridou, Elissavet
- Σιγγιρίδου, Ελισσάβετ
Subjects
dc:subject × 8Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/51397