Big data and artificial intelligence

Description

According to M. Minsky, Artificial Intelligence is the science of making machines do things that would require intelligence if done by men. Although the term was invented in a workshop at Dartmouth College in 1956 and has suffered from several ``winters'' since then, it has recently regained interest among others thanks to the application of machine learning to image recognition, automatic car driving, games and recommendation systems analyzing big data.

NADI offers a broad expertise in Artificial Intelligence. Number of its members are participating to AI4Belgium initiative, which joins together academic experts and administrations and companies' representatives to develop AI applications and researchers in Belgium. To cite a few key expertise, E. Tuci conducts research in the aim of allowing bio-inspired robots to operate in a complex environment and to learn from their experience in an autonomous way. He designs control mechanisms underpinning complex behavioural, social, cognitive and communication capabilities and studies operational principles of cognition and learning in natural organisms. B. Frénay aims to make machine learning techniques more robust, interactive, interpretable and safe to use. Indeed, machine learning is now a widespread approach to solve many data intensive problems, but existing tools are sensitive to anomalies, hard to understand, hard to control and provide insufficient guarantees on their behaviour. He is also working with colleagues in NADI about AI in education and AI legislation. W. Vanhoof develops techniques that allow to automatically verify whether a given software implements a particular algorithm. These techniques have several applications, ranging from program comprehension over plagiarism detection and malware detection to advanced analyses and optimizations such as the automatic detection of parallelization strategies. I. Linden explores how business intelligence platforms can be developed to extract useful information from organisations. She also develops new ways to model, acquire and manipulate weakly structured data such as texts and expert knowledge. J.-M. Jacquet designs new programming methodologies to conceive programs by declaring what has to be solved, and not how to solve the problem at hand. This line of research has been supported by Walloon Region projects to produce e.g. ExpeSurf, an expert system in multi-layer engineering, and Seplans, an expert system in estate planning. He is also interested in the design of complex knowledge representation systems to model socio-technological systems. G. Perrouin studies how software testing techniques can be applied to AI algorithms to make them safer to use. He is also interested in the applications of AI to complex software systems. J.-N. Colin explores the application of reinforcement learning to smart honeypots.

From a human science point of view, A. de Streel and H. Jacquemin have investigated the Robots and AI legal framework. From that work, they have edited a reference book in 2018. The concept of ``Algorithmic Governmentality'' has been invented and thoroughly studied by A. Rouvroy. Together with N. Grandjean, J. Grosman, C. Lobet-Maris and Y. Poullet, she is studying the ethics of Artificial Intelligence. In his thesis, L. Costa has analyzed how privacy induces a new legal approach of the emerging digital technologies. In addition, C. de Terwangne, J. Herveg, B. Michaux, Y.Poullet and A. de Streel study data protection, in particular in the light of GDPR, liability, intellectual property and competition laws. M. Lognoul, A. de Streel and B. Michaux study the explainability of AI from a legal perspective. W. Hammedi studies the application of AI techniques to customer experience. A. Castiaux analyses the opportunities offered by AI to develop innovation, the role of innovation eco-systems in its development, and the impact of AI on organisational and societal changes.

Recently, as examples of interdisciplinary research, B. Frenay and Y. Poullet have been asked to draft, at the Consultative Committee Convention n° 108 request, new Council of Europe recommendations about Profiling at the AI age. An extended report on the topic will be published in the next future. NADI is also developing a global expertise on the development of AI within public administration.

Key publications

• A. Bibal and B. Frenay. Learning Interpretability for Visualizations using Adapted Cox Models through a User Experiment. 2016 NIPS Workshop on Interpretable Machine Learning in Complex Systems. Barcelona
• A. Bibal and B. Frénay. Interpretability of Machine Learning Models and Representations: an Introduction. In 24th European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning. Bruges, pp. 77-82, 2016.
• A. Bibal, M. Lognoul, A. de Streel, B. Frenay. Implementing Legal Requirements on Explainability in Machine Learning Artificial Intelligence and Law, 2020
• C. Colot, I. Linden and P. Baecke. A Survey on Mobile Data Uses. International Journal of Decision Support System Technology, 8(2), 29-49, 2016.
• L. Costa, Virtuality and capabilities in a World of Ambient Intelligence, Thesis defended at Namur (2015), ~Springer International Publishing, 2016.
• B. Dumas, B. Frénay and J. Lee. Interaction and User Integration in Machine Learning for Information Visualisation. in ESANN 2018 proceedings, European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning, pp. 97-104, Bruges, 2018.
• B. Frénay and B. Hammer. Label-noise-tolerant classification for streaming data. In Proc. International Joint Conference on Neural Networks, IJCNN 2017, pp. 1748-1755, 2017.
• M. Gianni, K. Gotzamani and I. Linden. How a BI-wise responsible integrated management system may support food traceability. International Journal of Decision Support System Technology, 8(2), 1-17, 2016.
• J.-B. Hubin, H. Jacquemin and B. Michaux (ed.), Le juge et l'algorithme : juges augmentés ou justice diminuée, Coll. du Crids n° 46, Bruxelles, Larcier, 2019, 301 p.
• J.-M. Jacquet, I. Linden, and M.-O. Staicu. Blackboard Rules: from a Declarative Reading to its Application for Coordinating Context-aware Applications in Mobile Ad Hoc Networks. Science of Computer Programming, 115-116: 79-99, 2016.
• H. Jacquemin. Comment lever l'insécurité juridique engendrée par le recours à l'intelligence artificielle lors du processus de formation des contrats, Droit, normes et libertés dans le cybermonde, Liber amicorum Yves Poullet, Bruxelles, Larcier, 2018, pp. 141-172.
• H. Jacquemin and J.-M. Van Gyseghem. Le big data en matière d'assurance à l'épreuve du RGPD, Bull. Ass., dossier 2017, Data Protection : l'impact du GDPR en assurances, pp. 233-260.
• R. Marion, A. Bibal and B. Frénay. BIR: A Method for Selecting the Best Interpretable Multidimensional Scaling Rotation using External Variables', Neurocomputing, vol. 342, pp. 83-96, 2019.
• M. Mesnard, E. Payet, and W. Vanhoof, Towards a framework for algorithm recognition in binary code. In Principles and Practice of Declarative Programming, 2016. ACM Press.
• A. Narayan, E. Tuci, F. Labrosse, M.H.M. Alkilabi, A Dynamic Colour Perception System for Autonomous Robot Navigating on Unmarked Roads, Neurocomputing (Elsevier), Vol. 275, pp. 2251-2263, 2018.
• G. Perrouin, M. Acher, M. Cordy, X. Devroey. Proceedings of the 1st International Workshop on Machine Learning and Software Engineering in Symbiosis, MASES@ASE 2018, Montpellier, France, September 3, 2018. ACM 2018
• Y. Poullet. Le RGPD face à l'intelligence artificielle, Coll. du CRIDS n°49, Bruxelles, Larcier, 2020.
• A. Rouvroy and Y. Poullet. Le droit de la responsabilité des acteurs de l'intelligence artificielle, Colloquium organized by the UCLille, September 6 and 7, 2020.
• A. de Streel and H. Jacquemin (eds). L'intelligence artificielle et le droit, Coll. du CRIDS n°41, Bruxelles, Larcier, 2017.
• M. Vu and B. Frénay. User-steering Interpretable Visualization with Probabilistic Principal Components Analysis. in ESANN 2019 - Proceedings, 27th European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning., pp. 349-354, 2019.

Key projects

• ARIAC by DigitalWallonia4.ai (2021 - …)
• BEM : Business Event Manager, Walloon project on workflow reconstructions (2010-13)
• EFFaTA-MEM : Evocative Framework for Text Analysis - MEdiality Models (2017 - …)
• EOS VeriLearn : Verifying Learning Artificial Intelligence Systems (2017 - …)
• DIGI4FED: Use of AI and Big Data to fight tax and social fraud (2020-2022)
• SEPLANS : Expert System in Estate Planning, Walloon project on estate planning with AI (2007-15).

Contact: Benoît Frénay