Johny MATAR : "Optimization of machine learning techniques for genomic clustering""

Abstract: In the field of bioinformatics, clustering recently appeared to be a very efficient technique for sequence analysis. While greedy and hierarchical algorithms are used in the majority of the available tools, spectral clustering was recently introduced as a new stakeholder in this field. Spectral clustering is an efficient technique for well-separated sequence clustering and GMM's (Gaussian Mixture Models) are often able to cluster overlapping groups given an adequately designed embedding. Yet, the available clustering tools, for biological sequences, present
many drawbacks especially that i- the most widely used ones require an accurate choice of a non-obvious identity or similarity threshold, ii- most of them are not designed to cluster potentially divergent sequences, and iii- the recent one that relies on the spectral clustering technique, and that does not require any user intervention or prior knowledge about the input sequences, is so slow and was not enough tested and validated.

Jury Composition :

M. Christophe GUYEUX Full professor Université Bourgogne Franche-Comté PhD Director
M. Rony DARAZI Full professor Université Antonine Reportur
M. Pierre SPITERI Full professor  Ecole Nationale Supérieure d’Electrotechnique, d’Electronique, d’Informatique, d’Hydraulique et des Télécommunications Reviewer
M. Flavien VERNIER Maître de conférences Université Savoie Mont Blanc Reviewer
M. Hicham EL KHOURY Associate Professor Université Libanaise PhD Co-supervisor
M. Jean-Claude CHARR Maître de conférences Université Bourgogne Franche-Comté PhD co-supervisor

Localization : UFR STGI 19 Avenue du Maréchal Juin 90016 BELFORT - Room : 108 - Bâtiment E