Kernel Variational Inference Flow for Nonlinear Filtering Problem
报告人： 史作强（清华大学）
时间：2025-11-14 10:00-11:00
地点：智华楼王选报告厅-101
Abstract:
We present a novel particle flow for sampling called kernel variational inference flow (KVIF). KVIF is capable to transform samples of initial distribution to a given target distribution like that in Stein variational gradient descent (SVGD). Furthermore, KVIF does not require the explicit formula of the target distribution which makes it applicable in filtering problems. Under proper assumptions, convergences of KVIF has theoretical assurance. Based on KVIF, we can construct filters with higher accuracy for posterior sampling. Extensive numerical experiments for comparison with other classical filters are also demonstrated to verify the performance of KVIF.
About the Speaker:
史作强，清华大学丘成桐数学科学中心长聘教授，北京雁栖湖应用数学研究院双聘研究员，主要研究方向为偏微分方程数值方法，图像处理和机器学习中的微分方程模型，非线性非平稳信号时频分析等，在ACHA，SIAM系列期刊，Advances in Mathematics，ARMA等国际知名学术期刊发表文章90余篇。
Your participation is warmly welcomed!
欢迎扫码关注北大统计科学中心公众号，了解更多讲座信息!
