AI RESEARCH
Flatness-Aware Stochastic Gradient Langevin Dynamics
arXiv CS.LG
•
ArXi:2510.02174v3 Announce Type: replace Flatness of the loss landscape has been widely studied as an important perspective for understanding the behavior and generalization of deep learning algorithms. Motivated by this view, we propose Flatness-Aware Stochastic Gradient Langevin Dynamics (fSGLD), a first-order optimization method that biases learning its dynamics toward flat basins while retaining the computational and memory efficiency of SGD and