---

TADALoT Project

The TADALoT Project is a project funded by the following call: Pack Ambition Recherche 2017, Région Auvergne-Rhône-Alpes.

TADALoT stands for:

• Tadalot: Anomaly Detection by Machine Learning using Atypical Losses and Transfer-learning
• Détection d’anomalies par apprentissage automatique avec des fonctions de coût atypiques et de l’apprentissage par transfert

publications

Partners

The project involves two funded academic partners:

• LabHC: Laboratoire Hubert Curien (UJM, Université Jean Monnet)
• CREATIS: Centre de Recherche en Acquisition et Traitement de l’Image pour la Santé (CNRS)

Focus of the project

The TADALoT project focuses on the machine learning foundations for imbalanced data and anomaly detection. While it is a pervasive issue in applications, settings involving imbalanced classes are generally under-studied in machine learning. The originality of TADALoT lies into this focus on imbalanced data in machine learning. More precisely, TADALoT has three main subjects of investigation: transfer learning, representation learning and optimization of atypical loss functions, all in a context of highly imbalanced data.