I was awarded with the "Prix Excéllence Doctorat" from the Université Jean Monnet
I defended my PhD thesis, entitled "A unified view of local learning: theory and algorithms for enhancing linear models"
"Communication-Efficient Decentralized Boosting while Discovering the Collaboration Graph" presented at MLPCD 2 NeurIPS Workshop
Adversarial Robustness Toolbox is now available on github
the paper L3SVMs, co-written with R.Emonet and M.Sebban, is now available on arxiv
the paper Lipschitz Continuity of Mahalanobis Distances and Bilinear Forms, co-written with R.Emonet and M.Sebban, is now available on arxiv
talk on c2lm at Lives ANR project kick-off at Univ. Marseille
L3SVMs: Landmark-based Linear Local Support Vector Machines
L3SVMs is a new local SVM method which clusters the input space, carries out dimensionality reduction by projecting the data on landmarks, and jointly learns a linear combination of local models.
- it captures non-linearities while scaling to large datasets
- it's customizable: projection function, landmark selection procedure, linear or kernelized