DUKe - Data User Knowledge
Interactive (Human in the loop) Machine Learning, Data mining or exploration
Responsable : Philippe LERAY : philippe.Lerayatls2n.fr
Responsable adjoint : Guillaume RASCHIA : Guillaume.Raschiaatls2n.fr
The DUKe (Data User Knowledge) research group, part of the LS2N laboratory (UMR CNRS 6004), University of Nantes, aims at proposing querying, mining and learning techniques that take into account
- data types (relational, spatial, graphical, temporal, stream, etc.),
- expert knowledge or user interactions through adapted visual supports
Moreover, the objective of the research team is to provide algorithms that
- consider user-related data, in particular issues like privacy, fairness and the value of personal data
- show good properties in terms of user interaction : anytime, incremental, fast, user-knowledge accounting or graphical representation
- allow to observe, analyze users' usage and then propose a user-system coevolution
Part of our work has, over the years, established connections with other research fields that handle large data sets, which analysis requires introduction of field-specific expert knowledge into models (biology, ethology, history, sociology, literature and education sciences).
Our current application fields are :
- Enterprise of the future
- Business Intelligence : customer relationship management, ...
- Manufacturing intelligence : predictive maintenance, ...
- Digital transformation : extraction and formalization of business knowledge
- Health of the future
- Personalized medicine : genome-wide association studies, ...
- Hospital of the future : treatment and immersion in virtual reality, ...
- Digital Humanities
- Learning Analytics : Mooc log mining, recommendation of teaching resources, ...
- Cultural heritage : helping historical resources annotation, recommendation of museographic resources ...
- Sociology, epistemology : cross-mining of digital traces, surveys and interviews
Thématiques de l'équipe
Our goal is to propose User-centered methods for Data management, Mining or Machine Learning. This involves four challenges
Les actus de l'équipe
L’équipe DUKe représentée à la conférence ICML 2018 !L’article « Clustering-based Random Forest of Predictive Trees for Extreme Multi-label Learning » signé Wisssam Siblini, Pascale Kuntz, F. Meyer vient d’être accepté à la conférence ICML 2018 (International Conference on Machine Learning, classée A*). Cette conférence aura lieu à Stockholm du…[ Lire la suite…]
Régis Gras, professeur émérite du LS2N (équipe DUKe), publie un nouvel ouvrage sur l’A.S.I. aux éditions CépaduèsLa théorie de l’Implication Statistique ou l’Invraisemblance du faux. Cépaduès, mai 2018. Auteur : Régis GRAS, Professeur émérite, LS2N/DUKe Préface : Pascale KUNTZ, Professeur, LS2N/DUKe Cet ouvrage récapitule les méthodes et les concepts fondés initialement par Régis Gras au cours…[ Lire la suite…]
