7/4/2018 · Particle-based variational inference methods (ParVIs) have gained attention in the Bayesian inference literature, for their capacity to yield flexible and accurate approximations. We explore ParVIs from the perspective of Wasserstein gradient flows, and make both theoretical and practical contributions. We unify various finite-particle approximations that existing ParVIs use, and recognize …
Understanding and Accelerating Particle-Based Variational Inference note that, from the optimization point of view, all existing ParVIs simulate the gradient ?ow , but no ParVI yet exploits the geometry of P 2(X) and uses the more appealing acceler-ated ?rst-order methods on the manifold P 2(X). Moreover,, Understanding and Accel erating Particle-Based Variational Inference Chang Liuy, Jingwei Zhuoy, Pengyu Chengz, Ruiyi Zhangz, Jun Zhuyx, Lawrence Carinzx y: Department of Computer Science and Technology, Tsinghua University z: Department of Electrical and Computer Engineering, Duke University x: Corresponding authors chang-li14@mails.tsinghua.edu.cn, Particle-based Variational Inference Methods (ParVIs): Represent the variational distribution q by particles update the particles to minimize KL p(q). More exible than classical VIs more particle-e cient than MCMCs. Related Work: Stein Variational Gradient Descent (SVGD) [3] simulates the gradient ow (steepest descending curves) of KL p on P H(X) [2]. The Blob and DGF methods [1] simulate the.
5/24/2019 · Understanding and Accelerating Particle-Based Variational InferenceChang Liu, Jingwei Zhuo, Pengyu Cheng, Ruiyi Zhang, Jun ZhuParticle-ba… Particle-based variational infer ence methods (ParVIs ) have gained attention in the Bayesian inference literature, for their capacity to yield flexible and ac… Proceedings of Machine Learning Research.
11/17/2019 · Understanding and Accel erating Particle-Based Variational Inference . Chang Liu , Jingwei Zhuo, Pengyu Cheng, Ruiyi Zhang, Jun Zhu, and Lawrence Carin.ICML 2019. [Paper & Appendix] [] []Introduction. The project aims at understanding the mechanism of particle-based variational inference methods (ParVIs e.g.
Stein Variational …
