Interpretable machine learning for real-world applications
Bitte benutzen Sie diese Kennung, um auf die Ressource zu verweisen:
https://doi.org/10.48693/431
https://doi.org/10.48693/431
Langanzeige der Metadaten
DC Element | Wert | Sprache |
---|---|---|
dc.contributor.advisor | Prof. Dr. Gordon Pipa | ger |
dc.creator | Stojanović, Olivera | - |
dc.date.accessioned | 2023-11-15T16:35:03Z | - |
dc.date.available | 2023-11-15T16:35:03Z | - |
dc.date.issued | 2023-11-15T16:35:03Z | - |
dc.identifier.uri | https://doi.org/10.48693/431 | - |
dc.identifier.uri | https://osnadocs.ub.uni-osnabrueck.de/handle/ds-2023111510066 | - |
dc.description.abstract | Recent severe failures of black box models in high stakes decisions have increased interest in interpretable machine learning. In this cumulative thesis, I discuss why black box machine learning models can fail and explain the potential of interpretable machine learning. After a general introduction into this topic, I present three examples of interpretable machine learning models that I developed for studies in different scientific fields: medicine, epidemiology, and remote sensing, which correspond to three publications that constitute the thesis. For each publication, I first explain the data context, the prediction task, why it is a challenging problem and how interpretable machine learning can help improve the outcome. Then, in each publication, I outline the methods, examine their performance, and discuss how interpretability adds to understanding the results and phenomena. The publications show that it is possible to design interpretable models that yield good predictions, but they also demonstrate that domain expertise and understanding of data context are crucial. The thesis concludes with an outlook on the future of interpretable machine learning. I argue that, especially when it comes to high stakes decisions, a better understanding of machine learning models will be crucial - also because new and future laws will increasingly regulate algorithmic decisions. | eng |
dc.rights | Attribution 3.0 Germany | * |
dc.rights.uri | http://creativecommons.org/licenses/by/3.0/de/ | * |
dc.subject | Interpretable Machine Learning | eng |
dc.subject | Machine Learning | eng |
dc.subject | Data Science | eng |
dc.subject.ddc | 500 - Natural sciences & mathematics | ger |
dc.title | Interpretable machine learning for real-world applications | eng |
dc.type | Dissertation oder Habilitation [doctoralThesis] | - |
thesis.location | Osnabrück | - |
thesis.institution | Universität | - |
thesis.type | Dissertation [thesis.doctoral] | - |
thesis.date | 2023-10-11 | - |
orcid.creator | https://orcid.org/0000-0001-9820-3479 | - |
dc.contributor.referee | Prof. Dr. Peter König | ger |
dc.contributor.referee | PD Dr. Georges Hattab | ger |
Enthalten in den Sammlungen: | FB08 - E-Dissertationen |
Dateien zu dieser Ressource:
Datei | Beschreibung | Größe | Format | |
---|---|---|---|---|
thesis_stojanovic.pdf | Präsentationsformat | 12,77 MB | Adobe PDF | thesis_stojanovic.pdf Öffnen/Anzeigen |
Diese Ressource wurde unter folgender Copyright-Bestimmung veröffentlicht: Lizenz von Creative Commons