04/23/2024 | News release | Distributed by Public on 04/23/2024 06:43
The European Research Council (ERC) has awarded its Advanced Grants for ground-breaking research. Three KU Leuven researchers are among this year's recipients: Luc De Raedt, Stefaan Poedts and Greet Van den Berghe.
ERC Advanced Grants are awarded to established researchers with a considerable academic track record. The grants are awarded for a five-year period and are usually worth up to €2.5 million.
Full professor Faculty of Engineering Science; Department of Computer Science; manager of Leuven.AI
Neural networks that effortlessly generate texts, images, or videos. Various AI tools have quickly found their way to a broad audience. Behind the scenes, there is already talk of a new wave in AI that potentially possesses an even higher complexity: neurosymbolic AI (NeSy).
"Language models like ChatGPT are trained on vast sets of data, but they have no notion of logic or probability. As a result, the models are not reliable for all tasks," says computer scientist Luc De Raedt, director of Leuven.AI. "For example, ChatGPT can generate a coherent sequence of words and sentences, but it lacks the formal knowledge of subjects such as mathematics, physics, or chemistry to handle certain problems or topics correctly. The integration of symbolic reasoning with deep learning, that is the basis of neurosymbolic AI."
In the DeepLog project, Professor De Raedt wants to lay the foundations for such integration. "We will identify the building blocks for neurosymbolic AI: so-called primitives that combine the learning ability of neural networks with knowledge and probability-based reasoning techniques. Ultimately, we want to create a novel framework for enabling the development of NeSy systems and their applications."
Full professor Faculty of Science, unit of Plasma-astrophysics and chair of the Leuven Computational Modelling Centre
Astrophysicist Stefaan PoedtsJust like on Earth, there is also a weather forecast in space to predict solar winds and storms. Because the effects of solar winds and storms have a significant impact on both satellites and ground-based technological systems on Earth, it is important to make these predictions as accurately as possible and take measures if necessary. Due to the large distance from the sun and the different layers that make up the atmosphere, many different physical parameters influence the space weather, making the prediction very complex.
"As a solution, a separate model is created for each effect and each zone between the sun and Earth. These models are then linked together to create the space weather forecast," says Professor Stefaan Poedts. "Due to the concatenation of the different models, an enormous amount of computational power is required to perform the analysis and predict solar winds in terms of time and strength. To such an extent that the calculation with the most realistic models takes longer than the two days it takes for a solar storm to reach Earth."
The Open SESAME project focuses on the first link in the chain, which models the sources of space weather closest to the sun, namely the solar atmosphere. "By integrating more physical processes, we can not only optimize the first link but also all subsequent models. This way, we can predict space weather not only better but also faster and take measures against impending space weather, such as temporarily reducing electricity transmission and covering or temporarily shutting down measuring equipment in satellites. This can partially prevent damage, which could potentially amount to 2.7 trillion (2.7 * 1021) dollars globally in an extreme event."
Full professor Faculty of Medicine; head of the Clinical Department and the Laboratory of Intensive Care Medicine
Intensivist Greet Van den Berghe |Some patients require intensive care as a consequence of sepsis, complicated surgery, or extensive trauma. Over the years, their chances of survival have improved and most patients recover within a few days, but around ten percent of them subsequently require prolonged - sometimes for weeks or even months - intensive care. Many of these patients require mechanical ventilation, and suffer from lingering organ dysfunctions and muscle weakness.
Professor Greet Van den Berghe encounters this situation daily as an intensivist at UZ Leuven. "In these prolonged critically ill patients, we frequently observe two problems: muscle weakness and impaired function of the adrenal gland, for which there is currently no treatment without reciprocal off-target effects. It was long believed that these are two distinct clinical entities, but we hypothesize that there may be one common underlying mechanism. If we succeed in identifying this pathway in this research project, we have a promising new starting point to tackle both problems simultaneously."
Initially, this project involves experimental research making use of animal models and human tissue, but the ultimate goal is to set up a clinical study. "Based on preliminary data from our research, we envision a number of candidate compounds that, in case the central hypothesis of this project would be confirmed, could lead to an effective treatment. This would represent a major breakthrough for the patients in the first place, but also for intensive care services because the prolonged care of these patients has a major impact on the available capacity."
The European Research Council (ERC) funds ground-breaking and innovative projects by Europe's finest researchers through five types of grant:
Starting Grants
Consolidator Grants
Advanced Grants
Proof of Concept
Synergy Grants
up to €10 million for 6 years
for groups of 2 to 4 Principal Investigators
ERC grant holders at KU LeuvenYour ERC project at KU Leuven?