A0154
Title: Propagation of chaos for mean-field Langevin dynamics and its application to model ensemble
Authors: Atsushi Nitanda - A*Star / NTU (Singapore) [presenting]
Abstract: Mean-field Langevin dynamics (MFLD) is an optimization method derived by taking the mean-field limit of noisy gradient descent for two-layer neural networks in the mean-field regime. Recently, the propagation of chaos (PoC) for MFLD has gained attention as it provides a quantitative characterization of the optimization complexity in terms of the number of particles and iterations. Remarkable progress in a recent study showed that the approximation error due to finite particles remains uniform in time and diminishes as the number of particles increases. By refining the defective log-Sobolev inequality a key result from earlier work an improved PoC result is established for MFLD, which removes the exponential dependence on the regularization coefficient from the particle approximation term of the optimization complexity. As an application, a PoC-based model ensemble strategy with theoretical guarantees is proposed.
