On Friday, January 19th, at 11:35 in room S6, we will have three short presentations:
- Ahmad Dabaja, new PhD student, will introduce himself and his work entitled Designing incentives for effective federated learning in a distributed system
- Khaoula Otmani, master student, will present her work entitled Byzantine-Robust Federated Learning via Shapley Value.
Abstract: In Federated Learning systems, the repetitive communication between server and clients leaves room for attacks aimed at compromising the integrity of the global model. I will present a defence strategy based on Shapley Value that detects and prunes malicious clients from the learning process. - Lucas Potin, PhD student, will present his work entitled Analyse de sous graphes pour identifier des phénomènes de corruption dans les marchés publics
Abstract: Création d’une méthode utilisant les sous-graphes discriminants afin de classer une collection de graphes parmi 2 labels. Application sur un jeu de données de graphes associés aux marchés publics afin de détecter un potentiel de fraude .Comparaison des différentes métriques de discriminance possible par rapport a d autres méthodes de la littératures ( valeurs de shapley, random Forest…)