9 papers
Muon Is Theoretically Wrong For Convolutions, But Empirically Effective
Thibaut Boissin, Thomas Massena, Mathieu Serrurier +1
Muon, an optimizer known for its efficiency, has a clear interpretation for matrix-valued updates, but convolutional kernels are stored as four-dimensional tensors. Standard implem…
LiST: Lipschitz Scaling Training for Robust and Calibrated Neural Networks
Arthur Chiron, Franck Mamalet, Thomas Massena +2
While accuracy, robustness, and calibration are all essential for reliable neural networks, they are often studied separately; developing models that satisfy all three simultaneous…
LipSSD: Lipschitz-Constrained Single-Shot Detection for Adversarially Robust Object Detection
Vincent Lébé, Yannick Prudent, Corentin Friedrich +3
Object detectors have many applications in safety-critical systems, but they are known to be sensitive to worst-case perturbations such as adversarial attacks, which limits their a…
From SGD to Muon: Adaptive Optimization via Schatten-p Norms
Thomas Massena, Corentin Friedrich, Mathieu Serrurier
Modern optimizers, like Muon, impose matrix-wise geometry constraints on their updates. These matrix-wise constraints can be unified under Linear Minimization Oracle (LMO) theory.…
Fast and Flexible Robustness Certificates for Semantic Segmentation
Thomas Massena, Corentin Friedrich, Franck Mamalet +1
Deep Neural Networks are vulnerable to small perturbations that can drastically alter their predictions for perceptually unchanged inputs. The literature on adversarially robust De…
Turbo-Muon: Almost-Orthogonal Pre-Conditioning for Fast Muon Updates
Thibaut Boissin, Thomas Massena, Franck Mamalet +1
Orthogonality-based optimizers, such as Muon, have recently shown strong performance across large-scale training and community-driven efficiency challenges. However, these methods…
Efficient Robust Conformal Prediction via Lipschitz-Bounded Networks
Thomas Massena, Léo andéol, Thibaut Boissin +4
Conformal Prediction (CP) has proven to be an effective post-hoc method for improving the trustworthiness of neural networks by providing prediction sets with finite-sample guarant…
An Adaptive Orthogonal Convolution Scheme for Efficient and Flexible CNN Architectures
Thibaut Boissin, Franck Mamalet, Thomas Fel +3
Orthogonal convolutional layers are valuable components in multiple areas of machine learning, such as adversarial robustness, normalizing flows, GANs, and Lipschitz-constrained mo…
DP-SGD Without Clipping: The Lipschitz Neural Network Way
Louis Bethune, Thomas Massena, Thibaut Boissin +6
State-of-the-art approaches for training Differentially Private (DP) Deep Neural Networks (DNN) face difficulties to estimate tight bounds on the sensitivity of the network's layer…