Causal Inference of Extremes: Recursive Max-Linear Models Under Observational Noise

Buck, Johannes Ernst-Emanuel

Johannes Ernst-Emanuel Buck

If you experience problems opening the document, please try this link.

Original title:: Causal Inference of Extremes
Original subtitle:: Recursive Max-Linear Models Under Observational Noise
Translated title:: Kausale Inferenz von Extremen
Translated subtitle:: Rekursive Max-Lineare Modelle unter Beobachtungsrauschen
Author:: Buck, Johannes Ernst-Emanuel
Year:: 2023
Document type:: Dissertation
Faculty/School:: TUM School of Computation, Information and Technology
Advisor:: Klüppelberg, Claudia (Prof. Dr.)
Referee:: Klüppelberg, Claudia (Prof. Dr.); Améndola Cerón, Carlos Enrique (Prof. Dr.)
Language:: en
Subject group:: MAT Mathematik
Keywords:: causal inference, directed acyclic graph, extreme value theory, graphical model, max-linear model, structural equation model, machine learning, identifiability
Translated keywords:: kausale Inferenz, gerichteter azyklischer Graph, Extremwerttheorie, grafisches Modell, max-lineares Modell, Strukturgleichungsmodell, machine learning, Identifizierbarkeit
TUM classification:: MAT 620
Abstract:: Recursive max-linear models have gained increasing interest for modeling extremal dependence. The causal structure of such models is represented by a Bayesian network, where node variables are defined as a max-linear function of parental node variables and an independent innovation. In this dissertation, we study recursive max-linear models under observational noise. Particularly, we investigate structural properties and causal inference of noisy max-linear models.
Translated abstract:: Rekursive max-lineare Modelle haben zunehmendes Interesse für die Modellierung extremaler Abhängigkeiten erlangt. Die kausale Struktur solcher Modelle wird durch eine max-lineare Funktion von übergeordneten Knotenvariablen und einer unabhängigen Innovation definiert. In dieser Dissertation untersuchen wir rekursive max-lineare Modelle unter Beobachtungsrauschen. Insbesondere betrachten wir strukturelle Eigenschaften und kausale Inferenz von verrauschten max-linearen Modellen
WWW:: https://mediatum.ub.tum.de/?id=1695755
Date of submission:: 24.01.2023
Oral examination:: 22.03.2023
File size:: 7745418 bytes
Pages:: 146
Urn (citeable URL):: https://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:bvb:91-diss-20230322-1695755-1-3
Last change:: 14.04.2023
BibTeX