第八届仿真方法和应用研讨会（第二轮通知）

主题：仿真与新质生产力

兰州 2025年6月27-28日

仿真技术广泛应用于社会经济、生产制造、交通运输、生态环境、军事等领域，是关系国民经济与国家安全的核心技术。为了加强国内外仿真学者的交流，建立仿真领域学界与业界的交流平台，国内外知名学者与业界专家发起召开仿真方法和应用研讨会。至今为止该研讨会已举办七届，此前会议分别在上海交通大学（2017年）、北京大学（2018年）、香港中文大学（深圳）（2019年）、复旦大学（2020年）、哈尔滨工业大学（2021年）、西南财经大学（2023年）、中国科学院大学（2024年）成功举行。

今年，由兰州大学主办的“第八届仿真方法和应用研讨会”将于6月27-28日在兰州召开。本届会议的主题是“仿真与新质生产力”，旨在为我国仿真理论研究赋能新质生产力实践应用贡献力量。

主办单位：兰州大学管理学院

会议时间：2025年6月27-28日

会议报名截止时间：2025年6月20日

会议地点：兰州富力万达文华酒店

组委会：（按姓氏拼音先后排名）

主   席：

洪   流     明尼苏达大学

胡建强     复旦大学

王康周     兰州大学

成   员：

陈昕韫     香港中文大学（深圳）

高思阳     香港城市大学

耿   娜     上海交通大学

胡照林     同济大学

姜广鑫     哈尔滨工业大学

李海东     中国科学院大学

刘彦初     中山大学

罗   俊     上海交通大学

彭一杰     北京大学

肖   辉     西南财经大学

张宁威     兰州大学

钟   颖     电子科技大学

特邀报告人：（按姓氏拼音先后排名）

陈昕韫                       香港中文大学（深圳）数据科学学院副教授

冯铭斌                       滑铁卢大学统计与精算系副教授

高思阳                       香港城市大学数据科学学院副教授

姜广鑫                       哈尔滨工业大学经济与管理学院教授

Raghu Pasupathy     普渡大学统计学系教授

彭一杰                       北京大学光华管理学院副教授

张新雨                       中国科学院数学与系统科学研究院教授

郑泽宇                       加利福尼亚大学伯克利分校工业工程与运筹学系副教授

会议日程：

【6月27日】 报到（地点：兰州富力万达文华酒店）

◆09:00-20:00  报到，领取会议材料

【6月28日】 特邀学者论坛报告（地点：兰州富力万达文华酒店）

◆08:30-08:40   主办方致辞

◆08:40-10:10   特邀报告（Raghu Pasupathy，彭一杰）

◆10:10-10:30   茶歇

◆10:30-12:00   特邀报告（张新雨，郑泽宇）

◆12:00-13:30   午餐（会议预定）

◆13:30-15:00   特邀报告（姜广鑫，高思阳）

◆15:00-15:20   茶歇

◆15:20-16:50   特邀报告（陈昕韫，冯铭斌）

◆17:00-18:00   圆桌论坛

报告主题：

特别致谢：

感谢期刊Computers & Operations Research（COR）及其主编Francisco Saldanha da Gama教授对本次仿真方法和应用研讨会的大力支持。为响应本次研讨会的主题，COR下开设了题目为“Stochastic Simulation and AI: Advancing the Frontiers of Decision-Making”的特刊论文征集。通过这次研讨会的交流，希望参会者之间能够充分交流，碰撞出思想的火花。我们将推荐并邀请各位参会学者将新的工作投稿至特刊。特刊征文链接为：https://www.sciencedirect.com/special-issue/322277/stochastic-simulation-and-ai-advancing-the-frontiers-of-decision-making。

会议注册方式：

（1）邮箱报名：发送会议回执（见附件）到邮箱summer_sim2025@163.com

（2）微信报名：扫描二维码

注册截止时间2025年6月20日23:59

会议费用：

所有参会人员食宿以及交通费自理，无需缴纳会议注册费。

会议推荐酒店：

兰州富力万达文华酒店：豪华大床房（470元/晚/间）、豪华双床房（470元/晚/间）。

前往会场方式：

会场地址：兰州市城关区天水北路52号兰州富力万达文华酒店

（1）兰州中川国际机场

机场大巴：乘坐机场大巴1号线à万达十字站下车à步行150米至会场（预计1小时40分钟，30元）

出租车：兰州机场à富力万达文华酒店（预计1小时，160元）

城际铁路：中川机场东站à兰州火车站（预计1小时，20元）à（后续参考方式3）

（2）兰州西站（高铁站）

出租车：兰州西站北出站口à富力万达文华酒店（预计35分钟，30元）

地铁公交车：兰州西站北广场地铁站（地铁1号线）à兰州大学地铁站D口（步行200米）à兰州大学公交站（16路公交车）à滩尖子公交站步行250米至会场（预计50分钟，6元）

（3）兰州火车站

出租车：兰州火车站à富力万达文华酒店（预计12分钟，11元）

公交车：兰州火车站公交站（16路公交车）à滩尖子公交站à步行250米至会场（预计40分钟，2元）

组委会联系方式：

姚王卿 电话：13972870190   邮箱：yaowq21@lzu.edu.cn

张宁威 电话：15515578987   邮箱：zhangnw@lzu.edu.cn

特邀报告简介：

特邀报告1：待确定

报告嘉宾：Raghu Pasupathy，普渡大学

报告摘要：待确定

报告嘉宾简介：Dr. Raghu Pasupathy is a Professor of Statistics at Purdue University, with a PhD from Purdue and a B.Tech from IIT Chennai. His research focuses on stochastic optimization, simulation, and uncertainty quantification. He has taught courses like Duality in Infinite-dimensional Optimization and Stochastic Processes, and has published extensively in top journals and conferences. Recent works include studies on stochastic trust-region algorithms and Frank-Wolfe recursion on probability spaces.

特邀报告2：跨场景预训练管理决策基础模型

报告嘉宾：彭一杰，北京大学

报告摘要：传统管理决策方法难适应新时期工业管理需求。跨场景预训练管理决策基础模型有望解决其中关键挑战：整合不同层次场景信息，优化各环节动态决策，打破信息孤岛和决策孤立，缓解我国工业软件“卡脖子”困境。提出新神经网络训练方法突破BP在并行化和泛化能力上的局限，与装备制造头部企业成立联合实验室解决AGV协同优化调度问题。

报告嘉宾简介：彭一杰，北京大学光华管理学院副教授。从事复杂系统随机仿真优化的方法论与理论研究，并将新方法应用于人工智能、金融工程与风险管理、健康医疗等领域。在Operations Research，INFORMS Journal on Computing，IEEE Transactions on Automatic Control 等期刊上发表高水平学术论文，曾获INFORMS Outstanding Simulation Publication Award。主持国家自然科学基金优青、原创探索、杰青项目等。

特邀报告3：Sufficiency-principled Transfer Learning via Model Averaging

报告嘉宾：张新雨，中国科学院

报告摘要：Domain aggregation in multi-source transfer learning faces a critical challenge: effectively integrating knowledge from heterogeneous sources while addressing statistical uncertainties. Existing methods rely on restrictive single-similarity assumptions (e.g., individual or combinatorial similarity) and often neglect practical variability, leading to suboptimal performance. To address these limitations, we propose a sufficiency-principled transfer learning framework that systematically balances model averaging and model selection during domain aggregation with unknown informative knowledge. The framework employs a sufficiency principle for quantifying transferable knowledge to eliminate the challenges of spurious correlation and perturbated evaluation. The unified model averaging algorithms accommodate both individual and combinatorial similarity regimes, and also has privacy-preserving mechanisms. Theoretically, we establish the asymptotic optimality, estimator convergence and asymptotic normality, for multiple source domain linear regression models with diverging parameters. Compared with existing results, we provide enhanced rate of converge for parameter of interest. Empirical validation through extensive simulations and an analysis of Beijing housing rental data demonstrates the statistical superiority of our framework over conventional domain aggregation methods. The proposed methodology extends beyond regression models, offering a generalizable paradigm for transfer learning in statistical decision theory.

报告嘉宾简介：张新雨，中科院数学与系统科学研究院研究员, 中国科学技术大学管理学院博士生导师。主要从事统计和计量经济学的理论和应用研究工作，具体研究方向包括模型平均方法及其在管理决策、经济预测、机器学习和生物医学等领域的交叉研究。担任期刊SCI期刊《JSSC》领域主编以及《系统科学与数学》等多个期刊的编委，曾获中国青年科技奖，先后主持国家自然科学基金杰出青年基金及其延续项目。

特邀报告4：COSIMLA: Combining Numerical Linear Algebra with Simulation for Large Markov Chains Computation

报告嘉宾：郑泽宇，加利福尼亚大学伯克利分校

报告摘要：This presentation introduces a fully integrated algorithm for combining simulation with numerical linear algebra, as a means of computing relevant quantities for Markov chains and Markov jump processes with large or infinite state space. Linear algebra is used to analyze the “center” of the state space, while simulation is used to estimate contributions from path excursions outside the “center”. The method yields consistent estimators and significant efficiency improvements.  This presentation includes joint work Peter Glynn, Alex Infanger and Yifu Tang.

报告嘉宾简介：郑泽宇，加州大学伯克利副教授、终身教职，任职于工业工程与运筹系（IEOR）、伯克利人工智能研究实验室（BAIR）。担任Operations Research与Naval Research Logistics期刊Associate Editor。

特邀报告5：Efficient Graph Sampling Methods under Partial Information

报告嘉宾：姜广鑫，哈尔滨工业大学

报告摘要：The structure of networks plays a crucial role in the assessment of systemic risk. However, the true structure of a network is often difficult to observe directly, which makes it essential to develop methods for sampling possible network configurations based on partial information, such as node degree sequences. In this work, we consider the problem of sampling bipartite graphs (e.g., bank-asset networks), directed graphs (e.g., interbank networks), and undirected graphs (e.g., social networks) under such partial information. We first derive exact bounds on the number of nodes that can be connected at each step, given a prescribed degree sequence. Building on these bounds, we propose a weighted-balanced random sampling algorithm for generating bipartite graphs that are consistent with the observed degrees. We then extend this algorithm to develop two additional random sampling algorithms for directed and undirected graphs, incorporating specific rules tailored to each case. In addition, we demonstrate the effectiveness of the proposed algorithm through numerical experiments.

报告嘉宾简介：姜广鑫，哈尔滨工业大学经济与管理学院教授。研究方向为随机模型与仿真、人工智能、金融工程与风险管理等，多篇文章发表在管理领域著名期刊 Operations Research，INFORMS Journal on Computing，IISE Transaction，IEEE Transactions on Automatic Control等。曾荣获中国系统工程学会青年科技奖、中国运筹学会金融工程与金融风险管理分会青年学者最佳论文奖（一等奖）、JORSC最佳论文奖、《管理世界》十佳论文奖等奖项。目前担任中国管理现代化研究会风险管理专业委员会副秘书长、中国运筹学会金融工程与金融风险管理分会副秘书长、中国管理现代化研究会理事、中国信息经济学会理事、管理科学与工程学会理事、中国计算机学会（CCF）计算经济专业组执委、期刊 APJOR、JORSC副编辑（Associate Editor）、Fundamental Research青年编委、《系统管理学报》领域编辑等职务。

特邀报告6：Ranking and Selection with Unknown Sampling Variances

报告嘉宾：高思阳，香港城市大学

报告摘要：We consider fixed-budget ranking and selection (R&S). This is a popular decision model in simulation optimization, with the goal of maximizing the probability of correctly selecting the best design among a finite set of alternatives within a given simulation budget. Existing R&S methods basically assume that the simulation noises of the designs follow normal distributions with known variances. We argue that the assumption of known variances is problematic, which will cause the derived sample allocation rules to underestimate the sampling uncertainty. In this research, we solve the fixed-budget R&S problem with unknown sampling variances. We follow the framework of the optimal computing budget allocation (OCBA) and establish optimality conditions, develop selection algorithms and characterize theoretical properties of the selection algorithms for this problem. In particular, Ryzhov (2016) made a conjecture about the optimality conditions of this problem. This research shows that this conjecture is partially correct, and fills in the missing parts of this conjecture.

报告嘉宾简介：Siyang Gao (高思阳) received the B.S. degree in Mathematics from Peking University in 2009 and the Ph.D. degree in Industrial Engineering from University of Wisconsin-Madison in 2014. Dr. Gao is the Associate Head and an Associate Professor with the Department of Systems Engineering, City University of Hong Kong. His research is devoted to simulation optimization, machine learning and their applications in healthcare management. His work has appeared in Operations Research, Manufacturing & Service Operations Management, INFORMS Journal on Computing, Production and Operations Management, IEEE Transactions on Automatic Control, etc. He is a recipient of the Best Conference Paper Award at the IEEE Conference on Automation Science and Engineering in 2019, Best Paper Award at the International Conference on Logistics and Maritime Systems in 2019, and the Best Young Faculty Paper Award at the International Research Conference on Systems Engineering and Management Science in 2018. Dr. Gao is currently serving as an Associate Editor of the journals IEEE Transactions on Automation Science and Engineering and Journal of Simulation.

特邀报告7：When Machine Learning Meets Importance Sampling: A More Efficient Rare Event Estimation Approach

报告嘉宾：陈昕韫，香港中文大学（深圳）

报告摘要：Motivated by telecommunication applications, we investigate the problem of rare event simulation for stationary distribution of queueing networks. A common approach of rare event simulation is importance sampling (IS). However, in general queueing network settings, it could be challenging to find a good importance distribution that can effectively reduce the variance. We identify that such challenge mainly lies in the high variance of path-dependent likelihood ratio. To address the issue, we propose a new method combining importance sampling with machine learning techniques. In detail, we apply machine learning method to estimate a marginal likelihood ratio, which has a smaller variance, via sampling from certain importance distribution. In addition to theoretic guarantees, the efficiency of our method is illustrated via a couple of queueing network examples. The talk is based on joint work with Ruoning Zhao at CUHK-SZ.

报告嘉宾简介：Xinyun Chen (陈昕韫) is currently an Associate Professor in the School of Data Science at The Chinese University of Hong Kong, Shenzhen. She received her Ph.D in Operations Research from Columbia University in 2014. Her research interests include applied probability, stochastic simulation and reinforcement learning. She has published papers in journals and conferences including Mathematics of Operations Research, Operations Research, Management Science and ICLR. She is currently serving on the editorial board of Operations Research, Journal of Applied Probability and Advances in Applied Probability.

特邀报告8：How Well can Machine Learn with Nosy Data? -- Deep Learning for High Dimensional Nested Simulation

报告嘉宾：冯铭斌，滑铁卢大学

报告摘要：Deep learning models have achieved remarkable success across a wide range of applications, from autonomous driving to generative AI. Their potential in financial and actuarial fields has captured significant interest among researchers, practitioners, and regulators. However, the lack of transparency and interpretability in these models has raised concerns about their reliability and resilience, which are critical factors for ensuring financial stability and fulfilling insurance obligations. In this study, we use stochastic simulation as a controlled data generation framework to explore fundamental questions about deep learning models, such as “How effectively do deep learning models learn from noisy data?" Our findings reveal intriguing insights into their behavior and capabilities. Building on these insights, we introduce an innovative nested simulation procedure that employs deep learning models as proxies to efficiently estimate tail risk measures for variable annuity hedging errors. This approach not only optimizes simulation resources by focusing on tail scenarios but also maintains transparency in the estimation process.

Our numerical experiments show that the deep learning proxies can accurately identify tail scenarios and can estimate tail risk measures just as accurately as traditional approaches but with less computational effort. Our work paves the way for more reliable and interpretable applications of deep learning in finance and insurance, offering a promising direction for future research and practical implementation.

报告嘉宾简介：Mingbin (Ben) Feng (冯铭斌) is an Associate Professor and Director of the Master of Actuarial Science Program at the University of Waterloo. He holds a PhD in Industrial Engineering and Management Sciences from Northwestern University. His research focuses on quantitative risk management, financial engineering, Monte Carlo simulation, and nonlinear optimization, with particular interest in efficient simulation algorithms for risk measurement and AI applications in actuarial science. Professor Feng has published in leading journals such as the INFORMS Journal on Computing (IJOC), ACM TOMACS, and the North American Actuarial Journal (NAAJ). He also serves as an Associate Editor for Naval Research Logistics. Actively engaged in the simulation community, he has served as a track co-chair, proceedings editor, publicity co-chair, and registration co-chair for the Winter Simulation Conference.

附件：第八届仿真方法和应用研讨会-参会回执（第二轮）.docx