Scott Pesme

16 papers A* 6A 1Journal 9
YearRankTypeTitle / Venue / Authors
2025 J jnl
A Theoretical Framework for Grokking: Interpolation followed by Riemannian Norm Minimisation.
CoRR
Etienne Boursier, Scott Pesme, Radu-Alexandru Dragomir
2025 J jnl
MAP Estimation with Denoisers: Convergence Rates and Guarantees.
CoRR
Scott Pesme, Giacomo Meanti, Michael Arbel, Julien Mairal
2024 A* conf
Implicit Bias of Mirror Flow on Separable Data.
NeurIPS
Scott Pesme, Radu-Alexandru Dragomir, Nicolas Flammarion
2024 J jnl
Implicit Bias of Mirror Flow on Separable Data.
CoRR
Scott Pesme, Radu-Alexandru Dragomir, Nicolas Flammarion
2024 A conf
Leveraging Continuous Time to Understand Momentum When Training Diagonal Linear Networks.
AISTATS
Hristo Papazov, Scott Pesme, Nicolas Flammarion
2024 J jnl
Leveraging Continuous Time to Understand Momentum When Training Diagonal Linear Networks.
CoRR
Hristo Papazov, Scott Pesme, Nicolas Flammarion
2023 J jnl
(S)GD over Diagonal Linear Networks: Implicit Regularisation, Large Stepsizes and Edge of Stability.
CoRR
Mathieu Even, Scott Pesme, Suriya Gunasekar, Nicolas Flammarion
2023 A* conf
(S)GD over Diagonal Linear Networks: Implicit bias, Large Stepsizes and Edge of Stability.
NeurIPS
Mathieu Even, Scott Pesme, Suriya Gunasekar, Nicolas Flammarion
2023 A* conf
Saddle-to-Saddle Dynamics in Diagonal Linear Networks.
NeurIPS
Scott Pesme, Nicolas Flammarion
2023 J jnl
Saddle-to-Saddle Dynamics in Diagonal Linear Networks.
CoRR
Scott Pesme, Nicolas Flammarion
2021 A* conf
Implicit Bias of SGD for Diagonal Linear Networks: a Provable Benefit of Stochasticity.
NeurIPS
Scott Pesme, Loucas Pillaud-Vivien, Nicolas Flammarion
2021 J jnl
Implicit Bias of SGD for Diagonal Linear Networks: a Provable Benefit of Stochasticity.
CoRR
Scott Pesme, Loucas Pillaud-Vivien, Nicolas Flammarion
2020 A* conf
On Convergence-Diagnostic based Step Sizes for Stochastic Gradient Descent.
ICML
Scott Pesme, Aymeric Dieuleveut, Nicolas Flammarion
2020 J jnl
On Convergence-Diagnostic based Step Sizes for Stochastic Gradient Descent.
CoRR
Scott Pesme, Aymeric Dieuleveut, Nicolas Flammarion
2020 A* conf
Online Robust Regression via SGD on the l1 loss.
NeurIPS
Scott Pesme, Nicolas Flammarion
2020 J jnl
Online Robust Regression via SGD on the l1 loss.
CoRR
Scott Pesme, Nicolas Flammarion