
- 296 pages
- English
- PDF
- Available on iOS & Android
About this book
Die vorliegende Arbeit beschĂ€ftigt sich mit der Entwicklung und Anwendung tiefer grafischer Modelle zur KausalitĂ€tsanalyse, insbesondere in multivariaten Zeitreihen. Ein Schwerpunkt liegt dabei auf der BerĂŒcksichtigung versteckter Störfaktoren und der Analyse nichtlinearer ZusammenhĂ€nge.Durch die Integration von Expertenwissen und den Einsatz von Proxy-Variablen können komplexe kausale Strukturen in verrauschten Daten, die ĂŒber nichtlineare KausalverknĂŒpfungen miteinander verbunden sind und von versteckten Störfaktoren beeinflusst sind, aufgedeckt werden. Die entwickelten Methoden ermöglichen nicht nur die SchĂ€tzung der IntensitĂ€t kausaler ZusammenhĂ€nge, sondern auch die Detektion und Attribution von Anomalien in multivariaten Zeitreihen.Ein weiterer Beitrag dieser Arbeit ist die Entwicklung einer neuen datengetriebenen Methode zur Partitionierung des Netto-Ăkosystem-Austauschs (engl. net ecosystem exchange). Durch die Anwendung eines tiefen Zustandsraummodells können die Tageswerte der Ăkosystematmung geschĂ€tzt werden, was fĂŒr das VerstĂ€ndnis des Klimawandels von groĂer Bedeutung ist.
Frequently asked questions
- Essential is ideal for learners and professionals who enjoy exploring a wide range of subjects. Access the Essential Library with 800,000+ trusted titles and best-sellers across business, personal growth, and the humanities. Includes unlimited reading time and Standard Read Aloud voice.
- Complete: Perfect for advanced learners and researchers needing full, unrestricted access. Unlock 1.4M+ books across hundreds of subjects, including academic and specialized titles. The Complete Plan also includes advanced features like Premium Read Aloud and Research Assistant.
Please note we cannot support devices running on iOS 13 and Android 7 or earlier. Learn more about using the app.
Information
Table of contents
- Chapter 1 Introduction
- 1.1 From causation to causality
- 1.2 Domain knowledge integration
- 1.3 Probabilistic graphical models
- 1.4 Variational inference
- 1.5 Challenges
- 1.6 Contributions of this thesis
- Chapter 2 Theoretical prerequisites
- 2.1 Fundamental concepts and definitions
- 2.2 Knockoffs
- 2.3 Causal models
- 2.4 Artificial neural networks
- 2.5 Deep graphical models
- 2.6 Time series analysis and forecasting
- 2.7 Time series anomaly detection
- 2.8 Time series anomaly attribution
- 2.9 Model identification
- Chapter 3 A Deep State Space Model for Partitioning Net Ecosystem Exchange
- 3.1 Introduction and motivation
- 3.2 Related work
- 3.3 Partitioning NEE using a DeepState
- Chapter 4 Experiments for Partitioning Net Ecosystem Exchange
- 4.1 FLUXNET dataset
- 4.2 Learning a dynamical model of nighttime Reco
- 4.3 Evaluation of daytime Reco forecasts
- 4.4 Obtaining GPP
- Chapter 5 Time Series Anomaly Attributionusing Counterfactual Reasoning
- 5.1 Introduction and motivation
- 5.2 Related work
- 5.3 Detecting anomalous intervals
- 5.4 Counterfactual interval replacement
- Chapter 6 Experiments for Time Series Anomaly Attribution using Counterfactual Reasoning
- 6.2 Experimental setup and results
- Chapter 7 Nonlinear Causal Link Estimationunder Hidden Confounding
- 7.1 Introduction and motivation
- 7.2 Related work
- 7.3 Proposed methods
- Chapter 8 Experimental Evaluation ofCausal Link Estimation underHidden Confounding
- 8.1 Datasets
- 8.2 CCEVAE experiments
- 8.3 SCEVAE experiments
- Chapter 9 Conclusions
- 9.1 Summary and thesis contributions
- 9.2 Future work
- Appendix A Proof of Proposition 2.2.3
- Appendix B Additional attribution results
- Appendix C Causal link estimation
- Bibliography
- List of Own Publications
- List of Figures
- List of Tables
- Notation
- Acronyms