Luxembourg’s high-performance computers are set to unlock countless new opportunities for businesses and academia alike. Emmanuel Kieffer, research and computer scientist at the University of Luxembourg explains the potential impact of HPCs on finance.
Can you tell me a bit about your background and your area of expertise?
I have a PhD in computer science which was funded in 2015 by the FNR, which was mainly in mathematical modelling and game theory, specifically a field they call multi-level optimization. I work in the parallel computing and optimization team at the University of Luxembourg where my main mission revolves around providing advanced research support on optimization and AI algorithms for HPCs.
You recently gave a lecture on HPC usage in finance. What can you tell me about it?
I gave a lecture on how to use HPCs with A. approaches in finance. I received a grant from the EIB to study how to design recommitment strategies in Private Equity and my lecture was about the findings of this research.
Throughout the eight-month study, we used HPCs and machine learning to train and optimise recommitment strategies in order to help investors maintain and keep high allocation to private equity while minimising the risk of default. This is a highly complex problem that is impossible to solve without HPCs because regular computers lack the capabilities to simulate the necessary investment and investor behaviour. Furthermore, HPC allows you to perform a tremendous number of simulations increasing the statistical confidence of your results and better predict what would happen under certain market conditions.
“[…] once we have successfully democratised HPCs and with the new coming European Master for HPC we will see a positive economic impact and continued growth of our finance sector.”
Emmanuel Kieffer, research and computer scientist at the University of Luxembourg
What were the takeaways from this research project?
When we started this project and discussed it with some private equity experts, they told us that most private companies have no automatic way of designing and learning recommitment strategies. Our research findings are among the first academic research to suggest some automatic strategies designed with AI algorithms which can be used by the private equity sector.
The findings suggested that recommitment strategies can be trained with AI algorithms to maximise the exposure to private equity while minimising the risk of overinvestment.
How widespread is the use of HPCs in Luxembourg’s finance sector?
Currently, we are still in the early stages but more and more financial applications can be enhanced using HPC. Once you manage to translate that you open the door to new ways of analysing results and new ways of using machine learning to develop more accurate predictions.
So the focus is really on acquiring the competencies and skills to understand how you can take advantage of multiple and distributed large-scale resources. The exciting part of HPCs is that we continuously see new applications that can be greatly improved using HPC. It’s a very innovative field with so many possibilities that are yet to be uncovered. We notably think that it could be a game changer for developing new sustainable models (eg sustainable agriculture, improved water management, etc..).
How do you expect HPCs to impact the economy?
I think it’s really fantastic that we now have a national supercomputer. It will attract even more companies to Luxembourg because they will not need to invest directly in an HPC themselves and will benefit from highly skilled experts. We have an excellent infrastructure, are well positioned in Europe and have a dedicated team of HPC experts who will be able to guide these companies. I’m sure that in the coming years, once we have successfully democratised HPCs and with the new coming European Master for HPC we will see a positive economic impact and continued growth of our finance sector.
This article was first published in the Silicon Luxembourg magazine. Read the full digital version of the magazine on our website, here. You can also choose to receive a hard copy at the office or at home. Subscribe now.
