Search results

Abstract Deep learning has become an extremely important technology in numerous domains such as computer vision, natural language processing, and autonomous systems. As neural networks grow in size and complexity to meet the demands of these applications, the cost of designing and training efficient models continues to rise in computation and energy consumption. Neural Architecture Search (NAS) has emerged as a promising solution to automate the design of performant neural networks. However, conventional NAS methods often require evaluating thousands of architectures, making them extremely resource-intensive and environmentally costly.This thesis introduces a novel, energy-aware NAS pipeline that operates at the intersection of Software Engineering and Machine Learning. We present CNNGen, a domain-specific generator for convolutional architectures, combined with performance and energy predictors to drastically reduce the number of architectures that need full training. These predictors are integrated into a multi-objective genetic algorithm (NSGA-II), enabling an efficient search for architectures that balance accuracy and energy consumption.Our approach explores a variety of prediction strategies, including sequence-based models, image-based representations, and deep metric learning, to estimate model quality from partial or symbolic representations. We validate our framework across three benchmark datasets, CIFAR-10, CIFAR-100, and Fashion-MNIST, demonstrating that it can produce results comparable to state-of-the-art architectures with significantly lower computational cost. By reducing the environmental footprint of NAS while maintaining high performance, this work contributes to the growing field of Green AI and highlights the value of predictive modelling in scalable and sustainable deep learning workflows. Jury Prof. Wim Vanhoof - University of Namur, BelgiumProf. Gilles Perrouin - University of Namur, BelgiumProf. Benoit Frénay - University of Namur, BelgiumProf. Pierre-Yves Schobbens - University of Namur, BelgiumProf. Clément Quinton - University of Lille, FranceProf. Paul Temple- University of Rennes, FranceProf. Schin'ichi Satoh - National Institute of Informatics, Japan

Abstract Deep learning has become an extremely important technology in numerous domains such as computer vision, natural language processing, and autonomous systems. As neural networks grow in size and complexity to meet the demands of these applications, the cost of designing and training efficient models continues to rise in computation and energy consumption. Neural Architecture Search (NAS) has emerged as a promising solution to automate the design of performant neural networks. However, conventional NAS methods often require evaluating thousands of architectures, making them extremely resource-intensive and environmentally costly.This thesis introduces a novel, energy-aware NAS pipeline that operates at the intersection of Software Engineering and Machine Learning. We present CNNGen, a domain-specific generator for convolutional architectures, combined with performance and energy predictors to drastically reduce the number of architectures that need full training. These predictors are integrated into a multi-objective genetic algorithm (NSGA-II), enabling an efficient search for architectures that balance accuracy and energy consumption.Our approach explores a variety of prediction strategies, including sequence-based models, image-based representations, and deep metric learning, to estimate model quality from partial or symbolic representations. We validate our framework across three benchmark datasets, CIFAR-10, CIFAR-100, and Fashion-MNIST, demonstrating that it can produce results comparable to state-of-the-art architectures with significantly lower computational cost. By reducing the environmental footprint of NAS while maintaining high performance, this work contributes to the growing field of Green AI and highlights the value of predictive modelling in scalable and sustainable deep learning workflows. Jury Prof. Wim Vanhoof - University of Namur, BelgiumProf. Gilles Perrouin - University of Namur, BelgiumProf. Benoit Frénay - University of Namur, BelgiumProf. Pierre-Yves Schobbens - University of Namur, BelgiumProf. Clément Quinton - University of Lille, FranceProf. Paul Temple- University of Rennes, FranceProf. Schin'ichi Satoh - National Institute of Informatics, Japan

BNAIC/BeNeLearn 2025 will be held at the University of Namur under the auspices of the Belgian-Dutch Association for Artificial Intelligence (BNVKI) and the Dutch Research School for Information and Knowledge Systems (SIKS). The conference aims at presenting an overview of state-of-the-art research in artificial intelligence and machine learning in Belgium, The Netherlands, and Luxembourg. More information and registration

Hepatocellular carcinoma is the most common primary liver cancer. Unfortunately, this tumor still has a high mortality rate due to the lack of effective treatments for its most advanced or poorly localized forms. As part of a partnership with the CHU UCL Namur - site de Godinne and with the support of Roche Belgium, researchers in the Department of Biomedical Sciences are trying to understand why liver tumor cells are so resistant to treatment, and to identify therapeutic alternatives to better target them.

Abstract Deep learning has become an extremely important technology in numerous domains such as computer vision, natural language processing, and autonomous systems. As neural networks grow in size and complexity to meet the demands of these applications, the cost of designing and training efficient models continues to rise in computation and energy consumption. Neural Architecture Search (NAS) has emerged as a promising solution to automate the design of performant neural networks. However, conventional NAS methods often require evaluating thousands of architectures, making them extremely resource-intensive and environmentally costly.This thesis introduces a novel, energy-aware NAS pipeline that operates at the intersection of Software Engineering and Machine Learning. We present CNNGen, a domain-specific generator for convolutional architectures, combined with performance and energy predictors to drastically reduce the number of architectures that need full training. These predictors are integrated into a multi-objective genetic algorithm (NSGA-II), enabling an efficient search for architectures that balance accuracy and energy consumption.Our approach explores a variety of prediction strategies, including sequence-based models, image-based representations, and deep metric learning, to estimate model quality from partial or symbolic representations. We validate our framework across three benchmark datasets, CIFAR-10, CIFAR-100, and Fashion-MNIST, demonstrating that it can produce results comparable to state-of-the-art architectures with significantly lower computational cost. By reducing the environmental footprint of NAS while maintaining high performance, this work contributes to the growing field of Green AI and highlights the value of predictive modelling in scalable and sustainable deep learning workflows. Jury Prof. Wim Vanhoof - University of Namur, BelgiumProf. Gilles Perrouin - University of Namur, BelgiumProf. Benoit Frénay - University of Namur, BelgiumProf. Pierre-Yves Schobbens - University of Namur, BelgiumProf. Clément Quinton - University of Lille, FranceProf. Paul Temple- University of Rennes, FranceProf. Schin'ichi Satoh - National Institute of Informatics, Japan

Florentin Rochet, a professor of computer security at the University of Namur, specializes in applied cryptography and secure communications. Against the backdrop of rapid advances in artificial intelligence and open-source solutions, he shares his analysis of the current situation and offers his outlook for the future. 

The Faculty of Medicine is equipped with several cross-disciplinary research entities that also rely on collaboration with UCL Godinne University Hospital within the Namur Institute for Life Sciences (Narilis) research institute..

This is a significant honor for two members of the UNamur School of Medicine: Professor Charlotte Beaudart, who heads the "clinical research" track of the Master’s program in biomedical sciences, and Professor Jonathan Douxfils (School of Medicine, URPC – NARILIS) have just joined the College of Young Researchers of the Royal Academy of Medicine of Belgium.

After nine years at the helm, Professor Jean-Pierre Gillet has handed over the leadership of the Department of Biomedical Sciences to Professor Marielle Boonen. This is an opportunity to look back on the significant growth of this department, which now trains nearly 500 students each year. 

Extensive deliberations have been conducted to strengthen the fundamental research focus of the Master’s program in Biomedical Sciences at UNamur by integrating the curriculum from the Master’s program in Biochemistry and Molecular and Cellular Biology (BBMC). This ambitious project has led to the design of an entirely new program, conceived to fully leverage the complementary strengths of these two programs. 

As part of the Inforums and CyberExcellence seminars, we are pleased to welcome Dr. Ryan Wails (Georgetown University), who will speak to us about cybersecurity. You will find the abstract of his presentation and his biography below.No registration required.We hope to see many of you at this event!On the Interplay of Modern Traffic Analysis and Internet Censorship & Circumvention TechniquesIn this talk, I will review the current state of real-world Internet censorship and some tools that network users employ to circumvent censorship. Then, I’ll take a forward-looking view on how censors might incorporate modern ML-based traffic analysis techniques to block users, highlighting the need for stronger circumvention tools. Finally, I’ll discuss our new internet censorship evasion technique called Unidentified Protocol Generation (published at USENIX Security 2025), which is capable of evading detection by state-of-the-art traffic analysis.Author bio: Ryan Wails is a postdoctoral researcher at Georgetown University studying network privacy and security. He completed his PhD while working at the U.S. Naval Research Laboratory. This lab is known for originating the Tor network and establishing the PETs research community, which is now among the top international research communities in computer security. Ryan is a core contributor to the Tor project and has published in leading conferences on network simulators, privacy attacks, privacy-preserving measurements, path selection algorithms, censorship circumvention, and website fingerprinting. He has received several international distinctions, including Best Paper awards at top conferences and recognition at community events for his contributions.