Regularized Stein Variational Gradient Flow

被引：0

作者：

He, Ye ^{[1
]}

Balasubramanian, Krishnakumar ^{[2
]}

Sriperumbudur, Bharath K. ^{[3
]}

Lu, Jianfeng ^{[4
]}

机构：

[1] Georgia Inst Technol, Sch Math, 686 Cherry St, Atlanta, GA 30332 USA

[2] Univ Calif Davis, Dept Stat, 399 Crocker Lane,1 Shields Ave, Davis, CA 95616 USA

[3] Penn State Univ, Dept Stat, 314 Thomas Bldg, University Pk, PA 16802 USA

[4] Duke Univ, Math Dept, Box 90320,120 Sci Dr, Durham, NC 27708 USA

来源：

FOUNDATIONS OF COMPUTATIONAL MATHEMATICS | 2024年

关键词：

Wasserstein gradient flow; Stein variational gradient descent; Particle-based sampling; Convergence to equilibrium; Mean-field analysis; Reproducing kernel Hilbert space; Regularization; CONVERGENCE; DIFFUSION; KERNELS;

D O I：

10.1007/s10208-024-09663-w

中图分类号：

TP301 [理论、方法];

学科分类号：

081202 ;

摘要：

The stein variational gradient descent (SVGD) algorithm is a deterministic particle method for sampling. However, a mean-field analysis reveals that the gradient flow corresponding to the SVGD algorithm (i.e., the Stein Variational Gradient Flow) only provides a constant-order approximation to the Wasserstein gradient flow corresponding to the KL-divergence minimization. In this work, we propose the Regularized Stein Variational Gradient Flow, which interpolates between the Stein Variational Gradient Flow and the Wasserstein gradient flow. We establish various theoretical properties of the Regularized Stein Variational Gradient Flow (and its time-discretization) including convergence to equilibrium, existence and uniqueness of weak solutions, and stability of the solutions. We provide preliminary numerical evidence of the improved performance offered by the regularization.

引用

页数：59

共 50 条

[1] Stein Variational Gradient Descent as Gradient Flow
Liu, Qiang
ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 30 (NIPS 2017), 2017, 30
[2] Stein Variational Policy Gradient
Liu, Yang
Ramachandran, Prajit
Liu, Qiang
Peng, Jian
CONFERENCE ON UNCERTAINTY IN ARTIFICIAL INTELLIGENCE (UAI2017), 2017,
[3] Stein Variational Gradient Descent Without Gradient
Han, Jun
Liu, Qiang
INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 80, 2018, 80
[4] Grassmann Stein Variational Gradient Descent
Liu, Xing
Zhu, Harrison
Ton, Jean-Francois
Wynne, George
Duncan, Andrew
INTERNATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE AND STATISTICS, VOL 151, 2022, 151
[5] On the geometry of Stein variational gradient descent
Duncan, A.
Nusken, N.
Szpruch, L.
JOURNAL OF MACHINE LEARNING RESEARCH, 2023, 24
[6] Projected Stein Variational Gradient Descent
Chen, Peng
Ghattas, Omar
ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 33, NEURIPS 2020, 2020, 33
[7] Stein Variational Gradient Descent with Variance Reduction
Nhan Dam
Trung Le
Viet Huynh
Dinh Phung
2020 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN), 2020,
[8] Stein variational gradient descent with local approximations
Yan, Liang
Zhou, Tao
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING, 2021, 386
[9] Stein Variational Gradient Descent as Moment Matching
Liu, Qiang
Wang, Dilin
ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 31 (NIPS 2018), 2018, 31
[10] Stein Variational Gradient Descent with Multiple Kernels
Ai, Qingzhong
Liu, Shiyu
He, Lirong
Xu, Zenglin
COGNITIVE COMPUTATION, 2023, 15 (02) : 672 - 682

← 1 2 3 4 5 →