Unknown

Dataset Information

0

Stein Variational Gradient Descent with Matrix-Valued Kernels.


ABSTRACT: Stein variational gradient descent (SVGD) is a particle-based inference algorithm that leverages gradient information for efficient approximate inference. In this work, we enhance SVGD by leveraging preconditioning matrices, such as the Hessian and Fisher information matrix, to incorporate geometric information into SVGD updates. We achieve this by presenting a generalization of SVGD that replaces the scalar-valued kernels in vanilla SVGD with more general matrix-valued kernels. This yields a significant extension of SVGD, and more importantly, allows us to flexibly incorporate various preconditioning matrices to accelerate the exploration in the probability landscape. Empirical results show that our method outperforms vanilla SVGD and a variety of baseline approaches over a range of real-world Bayesian inference tasks.

SUBMITTER: Wang D 

PROVIDER: S-EPMC6923147 | biostudies-literature | 2019 Dec

REPOSITORIES: biostudies-literature

altmetric image

Publications

Stein Variational Gradient Descent with Matrix-Valued Kernels.

Wang Dilin D   Tang Ziyang Z   Bajaj Chandrajit C   Liu Qiang Q  

Advances in neural information processing systems 20191201


Stein variational gradient descent (SVGD) is a particle-based inference algorithm that leverages gradient information for efficient approximate inference. In this work, we enhance SVGD by leveraging preconditioning matrices, such as the Hessian and Fisher information matrix, to incorporate geometric information into SVGD updates. We achieve this by presenting a generalization of SVGD that replaces the <i>scalar-valued</i> kernels in vanilla SVGD with more general <i>matrix-valued</i> kernels. Th  ...[more]

Similar Datasets

| S-EPMC8563972 | biostudies-literature
| S-EPMC6611597 | biostudies-literature
| S-EPMC2933224 | biostudies-literature
| S-EPMC7039878 | biostudies-literature
| S-EPMC7402485 | biostudies-literature
| S-EPMC10542513 | biostudies-literature
| S-EPMC6626569 | biostudies-literature
| S-EPMC7714173 | biostudies-literature
| S-EPMC7821235 | biostudies-literature
| S-EPMC10705587 | biostudies-literature