The research activities carried out by the ADVANSE team are in the field of the analysis of large databases in order to extract new knowledge. They relate to the following three areas:
- Data mining (DM) which continues historical work with particular emphasis on the proposal of pattern mining approaches integrating either new dimensions such as, for example, the spatial dimension in the case of spatio-temporal patterns and trajectories or else the dimensions associated with the different types of arcs which may exist in multigraphs;
- Analytical Visualization (VA) which emphasizes analytical reasoning facilitated by interactive visual interfaces. Such interfaces, integrating various methods of information representation, interaction or even automatic knowledge extraction, aim to allow users to extract, from complex and / or heterogeneous data, information directly supporting analysis, planning and decision making;
- Machine learning (ML) which, in addition to the various works carried out on the definition of new machine learning approaches (eg detection of rare clusters, labeling of topics), focuses on taking into account small or very large data sets via traditional approaches (eg SVM, Gradient Boosting, active learning) or more recent (eg deep learning). Indeed, Deep Learning has shown its effectiveness for classifying data in large volumes or with large numbers of dimensions (e.g. images, sounds, texts). However, the explicability of the results, or even the definition of a good architecture remain very experimental exercises and constitute real challenges for the scientific community.
The ADVANSE team, in its three axes, is developing work on both theoretical and experimental bases to address the associated issues. The team is taken to cross these different axes for various projects.