ĚÇĐÄÍřŇł°ć

27 August 2024

On the 3rd of June, George Osipov at the Department of Computer and Information Science (IDA) successfully defended his thesis in the intersection of theoretical computer science and artificial intelligence (AI), with a focus on parameterized complexity of constraint satisfaction problems (CSPs).

image cut from thesis cover

Congratulations on your PhD! What was your background when you started your PhD?
"I have a bachelor's degree in computer science from Free University of Tbilisi in Georgia (where I'm from) and two years in a master's program in Mathematics at Ilia State University in Georgia with one Erasmus year in Uppsala."

What is it like to be a PhD student at IDA?
"My time as a PhD was quite pleasant overall. Despite having no prior research experience, close support from my supervisors helped me get up to speed quite quickly, and I have been enjoying research ever since. I was part of Wallenberg AI autonomous systems and software program (WASP) graduate school, which made the experience even more interesting with various courses around Sweden, summer and winter schools and international trips."

You have just finished your PhD. What do you want to do next?
"I am continuing to work on research directions that emerged in the later years of my PhD, now with the help of an International postdoc grant from the Swedish Research Council. It allows me to travel for two years outside Sweden, and my plan is to visit two universities in the UK: Oxford University and Royal Holloway University of London."

A summary of the thesis by George Osipov

Many real-world tasks can be phrased in terms of assigning values to variables subject to a set of constraints. This makes the Constraint Satisfaction Problem (CSP) a versatile modelling tool both in practice and in theory of computing. In applications, it is often impossible to satisfy all constraints, but there exist assignments that satisfy almost all constraints.

In his thesis, George Osipov has studied the complexity of solving almost satisfiable CSPs. Classical complexity theory strongly suggests that finding optimal assignments to CSPs is computationally intractable. However, by adopting the lens of parameterized complexity, many problems become tractable by leveraging the special structure of the instances. The thesis is at the intersection of theoretical computer science and AI and is not guided by an immediate application. One possible application is efficient data processing and automatic reasoning, which can be crucial for drawing insights from the drastically increasing amounts of data. An attractive feature of parameterized algorithms in this regard is that they scale favourablyto larger inputs.

To the thesis:

Read more about George Osipov

Doctoral studies in computer and information science

More about WASP at IDA

Organisation

Latest news from LiU

Två män, en kvinna.

Hard rock of the year with a touch of LiU voices

The choirs of ĚÇĐÄÍřŇł°ć have achieved a new musical milestone. At the 2026 Grammis Awards, Ghost was named Best Hard Rock/Metal – where the contribution from LiU’s choirs on the latest album has now been highlighted as part of the success.

kvinna som sitter ute pĂĄ campus valla.

Jeanne Cilliers is LiU’s Professor of Economic History

"Almost everything we experience today has historical parallels," says Jeanne Cilliers, new professor of economic history at LiU. She is interested in demographic processes such as marriage, fertility and mortality.

A man with glasses is looking at himself in the mirror.

Digital twin could reveal alcohol consumption in crime cases

Using a digital twin, it is possible to predict with greater precision than at present how much alcohol a person has consumed and at what time. The study was conducted by researchers at LiU and the Swedish National Board of Forensic Medicine.