Machine Learning Group - ULB

The Machine Learning Group (MLG), is a research unit of the Department of Computer Science of the Faculty of Science of ULB (Université Libre de Bruxelles, Brussels, Belgium),

The MLG targets machine learning and behavioral intelligence research with a focus on time series analysis, big data mining, causal inference, network inference, decision-making models and behavioral analysis with applications in data science, medicine, molecular biology, cyber security and cooperative social dynamics.

The MLG exploits all aspects of machine learning, artificial intelligence and computer research to provide critical support in areas such as geographic data mining, fraud detection, analysis of important (real) data, intelligent decision making, behavioral studies (economic, social and political), complex systems analysis and computational biology/bioinformatics - gene expression and cancer detection, computer-assisted medicine and theoretical research.

The MLG's area of research :

• Machine learning: research on new algorithms for selecting and learning characteristics applied to various application areas (IoT, fraud detection, precision medicine, prediction, interpretability, security).

• Behavioural intelligence: research on behavioural aspects of decision making using game theory (evolutionary) and simulations to study the origins of pro-social behaviour, the role of emotions in decision making and trust in artificial systems.

• Big data: seeking scalable solutions for the analysis of complex data (e.g. high throughput, high volume, high speed) from ICT and medicine

• Bioinformatics : research and development on medical and biological issues using machine learning, statistical and modeling approaches.

Examples of completed projects :

• Brussels MOBI-AID: Brussels MOBILity Advanced Indicators Dashboard

• FutureICT 2.0 - ICTs for Social Sciences

• DEFEATFRAUD: evaluation and validation of in-depth engineering and learning solutions for fraud detection

Competencies of the MLG :

• Supervised and unsupervised learning : learning models from labeled and unlabeled data using classification, regression and clustering techniques.

• Feature selection : reduce the number of characteristics in order to identify those that are most relevant for classification, regression, etc.

• Behaioural intelligence : analyze experimentally and theoretically the interactions between humans and AI in order to develop meaningful models.

• Game theory : the theory of strategic decision-making in situations of cooperation and non-cooperation.

• Intelligent decision making : how to adapt to past decisions and anticipate future choices of agents in strategic situations.

• Computational biology and bioinformatics : develop calculation methods and software tools for the analysis and interpretation of biological data.

• Evolutionary dynamics : agent learning by social imitation or survival of the fittest.

• Free software : development of publicly available software packages and sharing of code analysis through online repositories.

• Time series analysis : multi-step analysis and prediction of multivariate temporal data.