Trial lecture
“Universal Coalgebra for Dummies”
Time and place: June 4th, 2025 4:15 PM. Kristen Nygaards sal (5370), Ole-Johan Dahls hus.
Main research findings
When a researcher has to study a complex probabilistic model, she can specify it as a probabilistic program. This is useful when investigating the probabilities of critical failure in systems of, for example, network routing or autonomous robots. The key idea is to write out your distribution as a computer program. Probabilistic programming languages provide two characteristic features: random sampling from basic distributions and conditioning on observed events. The distributions expressible in a probabilistic language can be highly complex by additionally using programming language features such as assignments, if-then-else statements, and loops. To analyze the specified distribution, one important objective is to find the expected values, or outputs, of the probabilistic program. In this thesis, we use symbolic execution to find the expected outputs. The main idea of symbolic execution is that program execution works with symbolic expressions rather than values. Instead of following a single path, the objective of symbolic execution is to analyze all paths through a program. To make symbolic execution of probabilistic programs meaningful, the correspondence between symbolic and concrete semantics is clarified in this thesis. This is done by introducing a compositional semantics for symbolic execution for probabilistic programs. We show how to use it to compute the expected output values of probabilistic programs.
Adjudication committee
- Associate Professor Ana Sokolova, University of Salzburg, Austria
- Professor Benjamin Kaminski, Saarland University, Germany
- Professor Peter Csaba Ølveczky, University of Oslo, Norway
Supervisors
-
Professor Einar Broch Johnsen, University of Oslo, Norway
-
Professor Andrzej Wasowski, IT University of Copenhagen, Denmark
Chair of defence
- Professor Xing Cai, Department of Informatics, University of Oslo
Contact information at Department: Mozhdeh Sheibani Harat