64 citations
- Toulouse School of EconomicsFR13 papers
- Institut de Recherche en Informatique de ToulouseFR5 papers
- Institut de Mathématiques de ToulouseFR4 papers
- Adrian CollegeUS3 papers
- Toulouse School of Economics ResearchFR3 papers
- Discovery Air (Canada)CA2 papers
- Institut Universitaire de FranceFR2 papers
- Laboratoire Jean KuntzmannFR2 papers
- Royal University of Fine ArtsKH2 papers
- Thales (France)FR2 papers
- Université Grenoble AlpesFR2 papers
- British Geological SurveyGB1 paper
20 papers
Gradient extremals, talwegs, valleys, and directional alignment for generic gradient descent
Pascal Bégout, Jérôme Bolte, Thomas Mariotti +1
Gradient extremals are loci along which the gradient is an eigenvector of the Hessian. These objects provide a natural geometric framework connecting several notions, notably valle…
Convergence of optimizers implies eigenvalues filtering at equilibrium
Jerome Bolte, Quoc-Tung Le, Edouard Pauwels
Ample empirical evidence in deep neural network training suggests that a variety of optimizers tend to find nearly global optima. In this article, we adopt the reversed perspective…
On the Private Estimation of Smooth Transport Maps
Clément Lalanne, Franck Iutzeler, Jean-Michel Loubes +1
Estimating optimal transport maps between two distributions from respective samples is an important element for many machine learning methods. To do so, rather than extending discr…
Bilevel gradient methods and the Morse parametric qualification condition
Jérôme Bolte, Quoc-Tung Le, Edouard Pauwels +1
We introduce the Morse parametric qualification condition for bilevel programming. Generic semi-algebraic functions are Morse parametric in a piecewise sense. Thus, bilevel program…
Convergence rate of Euler-Maruyama scheme for McKean-Vlasov SDEs with density-dependent drift
Anh-Dung Le
In this paper, we study weak well-posedness of a McKean-Vlasov stochastic differential equations (SDEs) whose drift is density-dependent and whose diffusion is constant. The existe…
A second-order-like optimizer with adaptive gradient scaling for deep learning
Jérôme Bolte, Ryan Boustany, Edouard Pauwels +1
In this empirical article, we introduce INNAprop, an optimization algorithm that combines the INNA method with the RMSprop adaptive gradient scaling. It leverages second-order info…