Trial lecture
“Explainable AI: Foundations and Methods for Transparent Models”
Time and place: September 26, 2025 11:15 AM. Kristen Nygaards sal (5370), Ole-Johan Dahls hus.
Main research findings
Today's factories collect huge amounts of data from sensors and machines. This data is important for improving product quality, detecting problems early, and making production more efficient. But two major problems often stand in the way: much of the data is incomplete, and the way data is analysed is often unclear and hard to reuse.
This thesis explores how knowledge graphs, as a way of representing information like a map of facts, can help solve these problems. By using formal, structured models to describe data and how it is used, we make data analysis more transparent and easier to share even among non-experts.
The thesis also shows how knowledge graphs can be used to predict missing data. For example, in car manufacturing, we can estimate missing welding measurements by reasoning over the structure of the data and expert knowledge. We also apply this to recommender systems, such as those used in online shopping.
The result is a more explainable, reusable, and reliable way to use data in industry, which helps companies make smarter decisions with greater confidence.
Adjudication committee
- Associate Professor Anisa Rula, University of Brescia, Italyk
- Professor Jiao Lei, University of Agder, Norway
- Associate Professor Ana Ozaki, Department of Informatics, University of Oslo, Norway
Supervisors
-
Associate Professor Ahmet Soylu, University of Oslo, Norway
-
Professor Arild Torolv Soetorp Waaler, University of Oslo, Norway
-
Associate Professor Evgeny Kharlamov, University of Oslo, Norway
-
Senior Research Scientist Roman Dimitru, SINTEF, Norway
-
Associate Professor Baifan Zhou, OsloMet, Norway
Chair of defence
- Professor Xing Cai, Department of Informatics, University of Oslo
Contact information at Department: Mozhdeh Sheibani Harat