IMDEA Software

Iniciativa IMDEA

Inicio > Eventos > Charlas Invitadas > 2022 > Information-Theoretic Generalization Bounds for Stochastic Gradient Descent
Esta página aún no ha sido traducida. A continuación se muestra la página en inglés.

Gergely Neu

viernes 18 de febrero de 2022

11:00am Meeting room 302 & Zoom3 (pass: 5551337)

Gergely Neu, Research Assistant Professor, Universitat Pompeu Fabra, Barcelona, Spain

Information-Theoretic Generalization Bounds for Stochastic Gradient Descent


We study the generalization properties of the popular stochastic optimization method known as stochastic gradient descent (SGD) for optimizing general non-convex loss functions. Our main contribution is providing upper bounds on the generalization error that depend on local statistics of the stochastic gradients evaluated along the path of iterates calculated by SGD. The key factors our bounds depend on are the variance of the gradients (with respect to the data distribution) and the local smoothness of the objective function along the SGD path, and the sensitivity of the loss function to perturbations to the final output. Our key technical tool is combining the information-theoretic generalization bounds previously used for analyzing randomized variants of SGD with a perturbation analysis of the iterates.