Pierre Laforgue
About me
I am currently a postdoctoral researcher at the University of Milan, Italy. I work among the LAILA team led by professor Nicolò Cesa-Bianchi.
My recent research interests revolve around multitask online learning with applications to federated learning and privacy preservation.
I am also currently co-supervising two PhD theses:
Giulia Clerici, working on nonstationary bandits for music recommendation (co-supervised with N. Cesa-Bianchi)
Tamim El Ahmad, working on sketching algorithms to scale-up operator-valued kernel machines (co-supervised with F. d'Alché-Buc)
I hold a PhD in Machine Learning, prepared at Télécom Paris under the supervision of professors Florence d'Alché-Buc and Stephan Clémençon.
My dissertation focuses on Deep Kernel Representation Learning for Complex Data and Reliability Issues (manuscript, defense slides).
During my PhD, I have been awarded a research grant by the chair Good in Tech to study algorithms in the presence of selection bias.
Previously, I jointly graduated from ENSAE Paris and Master M2 MASH at ENS Cachan and Université Paris-Dauphine.
You can find more about my background on my LinkedIn or in my resume.
Research Interests
My research interests revolve around Learning Theory, and include more precisely:
Multitask Online Learning
Federated Learning and Privacy Preservation
Bandit Algorithms
Robust Learning and Median-of-Means
Kernel Methods and Sketching Algorithms
Statistical Learning and Sample Bias Issues
My list of publications is available here.
News
Sep. 23: our paper on multitask online kernel Ridge regression got accepted at NeurIPS (joint work with PG Sessa, N. Cesa-Bianchi and A. Krause) !
Sep. 23: our paper on debiasing methods for image databases got accepted at Journal of Nonparametric Statistics (joint work with S. Clémençon and R. Vogel) !
Aug. 23: our paper on sparsified sketches for Lipschitz kernel methods got accepted at TMLR (joint work with T. El Ahmad and F. d'Alché-Buc) !
|