The PhD defence will be in Abels utsikt - Niels Henrik Abels hus. The host of the session will moderate the technicalities while the chair of the defence will moderate the disputation.
Ex auditorio questions: the chair of the defence will invite the audience to ask questions ex auditorio at the end of the defence.
Trial lecture
February 28th, time: 10:15 am, Auditorium 2, Vilhelm Bjerknes hus.
Uncertainty Quantification in Explainable AI
Main research findings
Predictive modeling is central to many scientific disciplines and industrial applications. Tree-based models have become an essential class of prediction models in applied machine learning due to their flexibility, computational efficiency, and oftentimes high prediction accuracy. However, the algorithmic and non-parametric nature of these models also makes it challenging to apply these models in real-world applications. This thesis explores topics related to accuracy, interpretability, and uncertainty quantification in prediction tasks with tree-based models. The research is motivated by the use of machine learning in automated valuation models in the real estate sector, which are widely applied in both the private and public sector as a decision-support tool for questions of a financial nature. The thesis proposes several ways to increase the interpretability and uncertainty quantification in these models while still obtaining a high predictive accuracy, making the models more trustworthy and useful to practitioners. Due to the importance of the housing market in the broader economy, this could have a significant positive impact on the decision-making process by homeowners, financial institutions, and governmental agencies.
Adjudication committee
- Professor Henrik Boström, KTH Royal Institute of Technology
- Associate Professor Yushu Li, University of Bergen
- Professor Ingrid Kristine Glad, University of Oslo
Supervisors
- Associate Professor Johan Pensar, University of Oslo
- Professor Dag Einar Sommervoll, NMBU
- Associate Professor Ida Scheel, University of Oslo
Chair of defence
Head of Department Erlend Fornæss Wold.
Host of the session
Professor Ingrid Kristine Glad, University of Oslo