Stein-Log-Sobolev Inequalities for the Continuous Stein Variational Gradient Descent Method
B1 R3119
The rate of the convergence of the Stein Variational Gradient Descent method for sampling via the lens of PDEs and interacting particle systems will be discussed.
Overview
The Stein Variational Gradient Descent method is a variational inference method in statistics that has recently received a lot of attention. The method provides a deterministic approximation of the target distribution, by introducing a nonlocal interaction with a kernel. Despite the significant interest, the exponential rate of convergence for the continuous method has remained an open problem, due to the difficulty of establishing the related so-called Stein-log-Sobolev inequality. Here, we prove that the inequality is satisfied for each space dimension and every kernel whose Fourier transform has a quadratic decay at infinity and is locally bounded away from zero and infinity. Moreover, we construct weak solutions to the related PDE satisfying exponential rate of decay towards the equilibrium. The main novelty in our approach is to interpret the Stein-Fisher information, also called the squared Stein discrepancy, as a duality pairing between H−1(Rn) and H−1(Rn), which allows us to employ the Fourier transform. We also provide several examples of kernels for which the Stein-log-Sobolev inequality fails, partially showing the necessity of our assumptions.
Presenters
José A. Carrillo, Professor of the Analysis of Nonlinear Partial Differential Equations in the Mathematical Institute at the University of Oxford and Tutorial Fellow in Applied Mathematics at The Queen's College
Brief Biography
José A. Carrillo is currently Professor of the Analysis of Nonlinear Partial Differential Equations in the Mathematical Institute at the University of Oxford and Tutorial Fellow in Applied Mathematics at The Queen's College. He previously held academic positions at Imperial College London, Universitat Autònoma de Barcelona, and Universidad de Granada, where he did his PhD. He works on kinetic equations and nonlinear nonlocal diffusion equations. He has contributed to the theoretical and numerical analysis of PDEs, and their simulation in different applications such as granular media, semiconductors and collective behavior. His main scholarship contributions in Analysis of PDEs are in aggregation-diffusion equations: nonlinear Fokker-Planck type equations; the use of optimal transport techniques and entropy methods to analyse theoretically and numerically gradient-flow structures for PDEs and their singularities; the analysis of diffusive and kinetic models for self-organization, and their implications in mathematical biology, control engineering and global optimization.