Interview with Jan Kronqvist, member of the Department of Computing’s Computational Optimisation Group.
STEM for Britain is a major scientific poster competition and exhibition which has been held in Parliament since 1997 and is organised by the Parliamentary & Scientific Committee. Chaired by Stephen Metcalfe MP, its aim is to give members of both Houses of Parliament an insight into the outstanding research work being undertaken in UK universities by early-career researchers.
In this article, we interview Jan Kronqvist who is a member of the Department of Computing’s Computational Optimisation Group who participated in a STEM for Britain poster competition on 8th March 2021. Find out about his research interest and experience of the competition below.
What happened during the event?
Normally the event would be held in Parliament, but due to COVID-19, the event had to be organized online. The competitors are divided into five scientific fields: Biological and Biomedical Sciences, Chemistry, Engineering, Mathematical Sciences, and Physics. I was competing in the Mathematical Sciences category.
During the event, I gave a short presentation via Zoom and the meeting was attended by Members of Parliament and the Scientific Committee.
What is your research about?
My research is focused on mathematical optimization, and specific algorithms and techniques for finding the best possible solution to problems with both continuous and integer variables.
So, what are these optimization problems?
In our everyday life we constantly encounter situations where we have to choose between multiple alternatives and where we wish to choose the best alternative. For example, something as simple as which route to take while walking or more complex choices such as how to best schedule and operate a production process, how to most efficiently apply radiation therapy, or how to train and analyze AI systems. These are what we refer to as optimization tasks, where we want to choose the best solution (alternative).
Often there is a quantity or measure that we either want to make as big as possible (e.g., profit) or as small as possible (e.g., energy consumption). A mathematical optimization problem is simply the optimization task described in a mathematical form. To describe the optimization task, we use variables that represent the quantities we can change or discrete decisions (e.g., yes/decisions) and constraints that describe how variables are linked to each other.
What do you like most about this field of research?
I have always enjoyed problem-solving, and I greatly enjoyed trying to make things work better. As a teenager, I was interested in engines and this actually further boosted my interest in applied mathematics.
You could consider my research to be advanced problem-solving. I am mainly focusing on the mathematical and computational aspects of solving optimization problems, i.e., developing algorithms and software for solving optimization problems.
What are the potential benefits and challenges of your research?
The potential benefits are numerous. For example, by optimizing production processes or electric power generation it is possible to greatly reduce emissions and consumption of raw materials. I know researchers who are also using mixed-integer optimization for designing new molecules and materials with optimal properties. I have recently also been working on using mixed-integer optimization for AI systems, and here we use it for analyzing safety aspects. Mixed-integer optimization is a very versatile tool.
I would say that the main challenge is how to computationally solve optimization problems of real-world relevant size. Solving mixed-integer optimization problems can be very difficult. A lot of progress has been made, and there are many problems that we are able to solve surprisingly well. But there are also fairly small problems that we struggle to solve with today's solvers and algorithms. There is definitely a need for both further theoretical and practical research in optimization, to be able to obtain solutions to current and future problems.
What do you consider as your main contributions so far?
I have focused on a specific type of optimization problem that we call convex mixed-integer nonlinear optimisation, which contains a large variety of problems covering a wide range of practical applications. These problems have some nice mathematical properties, that I have used to develop several new algorithms.
The new algorithms offer more efficient ways to solve the problems, making some previously intractable problems solvable. As an example where the new algorithms and techniques really made a big difference; there were several problems that we were not able to solve by running the computer for a day, but by applying a reformulation technique that I developed we were able to solve the problems in a less than 10 seconds.
Based on the algorithms I have also developed a new open-source solver together with a colleague, which is considered as the most efficient tool for this type of problem, With the solver, I feel that my research has pushed the field a step forwards.
Developing a solver has been very interesting. Especially for mixed-integer optimization, it is not only about coming up with the most efficient algorithm. I would actually say that it is more about how you can efficiently combine different techniques and algorithms in an efficient way to best take advantage of their strengths.
After two years at Imperial, you’re now moving to KTH to take up a professorship. What are your plans for your new role?
I will continue working on mixed-integer optimization, and I have several exciting new ideas that I want to explore further.
How has your time at Imperial impacted your research career?
Imperial College is a great environment for research. The interactions with leading researchers create an inspiring environment, and there is a great environment for collaborations. My time at Imperial has given me a broader perspective and also helped me to expand my research into new directions.
Watch Jan's presentation below
Article text (excluding photos or graphics) © Imperial College London.
Photos and graphics subject to third party copyright used with permission or © Imperial College London.
Ahmed Abdullahi Idle
Department of Computing