differential equations 1finite-sample analysis 1kernel methods 1physics-informed learning 1reproducing kernel Hilbert space 1universal consistency 1
From the 1 of 2 linked papers with an AI index.
2 papers
stat.ML2026
PIKS: Universal Physics-Informed Kernel Methods
Joachim Bona-Pellissier, Giacomo Meanti, Matteo Santacesaria +1
The paper proposes Physics-Informed Kernel Methods (PIKS), a kernel-based approach that incorporates linear differential constraints into learning, proving universal consistency an…
cs.AI2026
Geometry-induced Regularization in Deep ReLU Neural Networks
Joachim Bona-Pellissier, François Malgouyres, François Bachoc
Neural networks with a large number of parameters often do not overfit, owing to implicit regularization that favors \lq good\rq{} networks. Other related and puzzling phenomena in…