Victor Bouvier, a PhD student under the supervision of Professor Céline Hudelot and funded by a CIFRE collaration with Sidetrade, has been the recipient of the best (student) paper award at the conference ECML/PKDD.

Victorcurrently works on making AI models more robust to changes in the data. In the paper "Robust Domain Adaptation: Representation, Weights and Inductive Bias", published at ECML/PKDD2020, Victor has developed a unified theoretical framework for reconciling two major lines of study of the literature.

Importantly, Victor shows that inductive bias of the AI (the set of assumptions an engineer can enforce into its AI) makes consistently the AI able to adapt to new situations. Various experiments demonstrate the effectiveness of this novel theoretical framework.