一周活动预告: 4.10-4.16
Adaptive Randomized Sketching for Dynamic Nonsmooth Optimization (Harbir Antil)
Continuous-in-time Limit for Bayesian Bandits (Professor Yuhua ZHU)
Numerical algorithms for inverse spectral problems （徐翔）
Inverse random scattering problems for wave equations （王旭）
Variational inverting network for statistical inverse problems of partial differential equations （贾骏雄）
Numerical studies of domain sampling methods for inverse boundary value problems by one measurement （王海兵）
Methods in Solving Multi-Agent Multi-Armed Bandits（Hao Jin）
复杂流体建模、分析、算法培训班
1. Adaptive Randomized Sketching for Dynamic Nonsmooth Optimization
Speaker: Harbir Antil (George Mason University)
Time: 2023-4-11, 16:00 CEST+0200 (Europe/Rome)
Registration and Source Link:
https://na-g-roms.github.io/seminars/Harbir_Antil_2023.html
Abstract:
Dynamic optimization problems arise in many applications including optimal flow control, full waveform inversion, and medical imaging, where they are plagued by significant computational challenges. For example, memory is often a limiting factor on the size of problems one can solve since the evaluation of derivatives requires the entire state trajectory. Additionally, many applications employ nonsmooth regularizers such as the
（灰色区域内上下滑动阅读全部内容）
2. Continuous-in-time Limit for Bayesian Bandits
Source Link:
https://hkumath.hku.hk/MathWWW/event/2023/NAS-ProfYuhuaZHU.pdf
3. Numerical algorithms for inverse spectral problems
报告人: 徐翔 （浙江大学）
报告时间: 2023-4.12 14:00
报告链接: 腾讯会议ID: 161 849 796
信息来源:
https://math.jlu.edu.cn/info/1555/14463.htm
报告摘要:
In this talk, we will discuss some recent progress on numerical algorithms for inverse spectral problem based upon trace formulas for some typical differential operators. Instead of inverting the map from spectral to coefficients directly, we propose a novel method to reconstruct the coefficients based on inverting a sequence of trace formulas which bridge the spectral and geometry information clearly in terms of a series of nonlinear Fredholm integral equations. Numerical experiments are presented to verify the validity and effectiveness of the proposed numerical algorithm. The impact of different parameters involved in the algorithm is also discussed.
（灰色区域内上下滑动阅读全部内容）
4. Inverse random scattering problems for wave equations
报告人: 王旭（中科院）
报告时间: 2023-4.12 14:00-15:00
报告链接: 腾讯会议ID: 836 122 148
信息来源:
http://math.xjtu.edu.cn/info/1089/12020.htm
报告摘要:
In this talk, inverse random scattering problems with a random source or potential will be introduced for different time-harmonic wave equations. The unknown random source or potential is assumed to be a generalized isotropic Gaussian random field with its covariance operator being a classical pseudo-differential operator. With information of the data observed in a bounded domain, the strength of the random source or potential, involved in the principal symbol of its covariance operator, is shown to be uniquely determined by a single realization of the magnitude of the wave field averaged over the frequency band with probability one.
（灰色区域内上下滑动阅读全部内容）
5. Variational inverting network for statistical inverse problems of partial differential equations
报告人: 贾骏雄（西安交通大学）
报告时间: 2023-4.13 10:00-11:00
报告链接: 腾讯会议ID: 650 107 514
信息来源:
https://math.jlu.edu.cn/info/1555/14462.htm
报告摘要:
For quantifying the uncertainties of the inverse problems governed by some partial differential equations (PDEs), the inverse problems are transformed into statistical inference problems based on Bayes' formula. Recently, infinite-dimensional Bayesian analysis methods have been introduced to give a rigorous characterization and construct dimension-independent algorithms. However, there are three major problems for current infinite-dimensional Bayesian methods: prior measures usually only behave like regularizers; complex noises are rarely considered; many computationally expensive forward PDEs need to be calculated for estimating posterior statistical quantities. To address these issues, we propose a general infinite-dimensional inference framework based on a detailed analysis of the infinite-dimensional variational inference method and the ideas of deep generative models that are popular in the machine learning community. Specifically, by introducing some measure equivalence assumptions, we derive the evidence lower bound in the infinite-dimensional setting and provide possible parametric strategies that yield a general inference framework named variational inverting network (VINet). This inference framework has the ability to encode prior and noise information from learning examples. In addition, relying on the power of deep neural networks, the posterior mean and variance can be efficiently generated in the inference stage in an explicit manner.
（灰色区域内上下滑动阅读全部内容）
6. Numerical studies of domain sampling methods for inverse boundary value problems by one measurement
报告人: 王海兵（东南大学）
报告时间: 2023-4.13 14:00-15:00
报告链接: 腾讯会议ID: 126 761 357
信息来源:
https://math.jlu.edu.cn/info/1555/14464.htm
报告摘要:
We consider an inverse boundary value problem for the Laplace equation, which discusses the reconstruction of an unknown target inside the background medium from one boundary measurement. We are interested in two domain sampling methods, i.e., the range test and no-response test. We study the convergence and numerical realizations of these methods. Some new techniques are proposed to set up efficient algorithms, which yield reasonably good numerical reconstructions. To demonstrate the performance of proposed algorithms, we show several numerical examples for different shapes of unknown targets with noisy measurement data. Some key ingredients of numerical implementations are discussed in detail.
（灰色区域内上下滑动阅读全部内容）
7. Methods in Solving Multi-Agent Multi-Armed Bandits
报告人: Hao Jin (PKU)
报告时间: 2023-4-14 16:00-17:00
报告地点: 腾讯会议ID: 723 1564 5542
信息来源: https://www.math.pku.edu.cn/kxyj/xsbg/tlb/informationsciences/148489.htm
报告摘要:
Recently, multi-armed bandit algorithms have been deployed in many realistic large-scale applications, i.e., ranking of search engines and acceleration of model selection. In these cases, the workload represented by the number of bandits |A| is simply too high to be handled by a single processor. A natural way out is to introduce several independent nodes that may take actions and observe rewards in parallel. We denote such problems as multi-agent multi-armed bandits.
In this talk, we introduce various methods for solving multi-agent multi-armed bandits, most of which are derived from classical solutions to single-agent multi-armed bandits. In addition to regret analysis for any participated node, we concentrate on the communication overhead of these algorithms. Moreover, the graph structure of the network in which independent nodes are connected greatly affects the design and analysis of proposed algorithms. Take T, K and N as the time horizon, number of bandits and number of participated agents. Recent works have achieved a communication overhead of O(\log(T)\log(K)) while enjoying a \sqrt{N} speed up for any participated node in finding the optimal arm.
（灰色区域内上下滑动阅读全部内容）
8. 复杂流体建模、分析、算法培训班
会议地点: 北京计算科学研究中心
会议时间: 2023 4.19-4.23
信息来源:
http://www.csrc.ac.cn/en/event/workshop/2023-03-31/116.html
注册投稿网址:
http://csrc-supercomputer.mikecrm.com/Te3tWT9
会议信息:
随着计算机快速发展，多种新型计算模式如雨后春笋般出现，挑战了传统数值计算方法，带来计算数学领域众多变革。新型计算模式在航空航天，科技发展，国防工业等领域日趋重要。中国工程物理研究院承担重要国防科技任务，新型计算模式在中物院科研任务中发挥了重要作用。北京计算科学研究中心作为中物院对外交流的窗口，大力支持相关领域国内外学者的相互交流，促进计算数学和计算力学更好地为国家的国防建设服务。
本研讨会旨在促进国内外学者交流最新研究进展，探讨新的计算模型。我们还将一起讨论计算数学新、旧计算模式如何更好的碰撞、融合，如何更好解决流体力学实际问题，如何更好应用服务于国防科技事业。会议同时帮助更多青年学者与高年级研究生了解新型计算模式，加强对国内外最新科研进展的了解，调动科研积极性与发展掌握解决实际问题能力。会议主题包括但不限于：新型计算模式、多介质流体力学、高精度保物理性质算法，最优输运算法等。
会议分为两个部分，4/19-20为期两天的机器学习基础短课，21-23为期三天的小型研讨会。欢迎青年学者与高年级研究生参加，会议注册费500元（可申请免费名额，请导师推荐），会议期间提供午餐。另外设置Poster环节，欢迎青年学者与高年级研究生投稿。
（灰色区域内上下滑动阅读全部内容）
我们是信息搬运工，
若活动临时有变化，
请以网站信息为准。
欢迎转发，点赞，在看。👇