召集人：吴建福（美国佐治亚理工大学工业与系统工程学院）、黄建华（香港中文大学（深圳）数据科学学院）、熊世峰（中国科学院数学与系统科学研究院）

时间：2024.08.25-2024.08.31

DTUQ会议日程

日程总览表

会议日程Day1 (8月26日，周一)

会议日程Day2 (8月27日，周二)

会议日程Day3 (8月28日，周三)

会议日程Day4 (8月29日，周四)

会议日程Day5 (8月30日，周五)

报告摘要与报告人简介：

集成物理先验的生成模型在AI4Science中的应用

于天舒

摘要：

AI4Science目前是一个非常前沿并且具有前景的研究领域，能够大大加速传统的物理、化学、材料科学的研究和发现进程。与传统的基于模型的科学研究不同，AI4Science更加强调数据，通过对观测数据的学习来挖掘潜在规律。近年来，生成式人工智（Generative AI）能展现出了对数据建模的强大力量，并被广泛的应用到了AI4Science领域。但是，如何将物理/化学的先验知识集成到生成式人工智能，形成更加符合物理规律的建模，一直是个难点。本讲座简要介绍生成式人工智能在AI4Science领域的历史、方法、应用场景，以及团队在集成物理/化学先验知识的生成式人工智能方面的一些探索。

个人简介：

于天舒博士现为香港中文大学（深圳）助理教授。研究领域有机器学习、组合问题优化、图学习和优化、循环神经网络、行列式点过程等。于博士主要研究兴趣涵盖多个机器学习和组合问题优化相关领域。他对利用机器学习解决传统组合问题，图学习和优化，以及在深度学习框架内寻求结构扩展特别感兴趣。循环神经网络，行列式点过程也是他开展的研究之一。于博士也是多个顶级会议（例如ICLR 2021, NIPS 2020, CVPR 2019-2021, ICCV 2019, ECCV 2020等）以及期刊（例如IEEE Transactions on Image Processing, IEEE Transactions on Neural Networks and Learning Systems, IEEE Transactions on Circuits and Systems for Video Technology, Pattern Recognition, Pattern Recognition Letters等）的审稿人。

基于小样本数据的复杂物理模型不确定性量化研究

王言金

摘要：

在复杂物理问题的研究中常因过高的试验成本难以获得足够的实验数据导致小样本问题，其难点在于数据集信息不足以描述原始问题的全部特征。小样本问题存在两个方面的难点：样本量小，难以获取其包含的物理信息；由于样本量小，样本质量难评估，其样本质量差异对评估结果会造成巨大的影响。高能炸药在国民经济和国防建设中发挥着重要的作用，其宏观物理建模均是唯象建模，其中准确的未反应炸药状态方程是进行炸药爆轰流场高保真计算和炸药装药冲击起爆模拟等研究的基础。冲击Hugoniot实验是研究未反应炸药的基准实验。首先，本报告针对Hugoniot实验小样本问题，从数据增强与样本质量评价两个角度发展了不确定度量化方法解决小样本问题，用于评估未反应炸药的物理模型参数不确定性，开展了未反应炸药的不确定度传播研究，结合物理认知分析了不确定度传播结果。其次，本报告基于未反应炸药物理模型的不确定性量化分析结果，开展了包含炸药物理模型、产物物理模型、反应率物理模型等模型的爆轰全模型的不确定度性量化研究，建立了自主学习的多保真代理模型方法，评估了基于层级实验的爆轰模型不确定度研究。

个人简介：

王言金，博士，北京应用物理与计算数学研究所特聘研究员。武汉大学计算数学博士；日本东京大学数理科学博士。作为负责人多次承担了国家自然科学基金委项目。2018年至今担任中国工业与应用数学学会不确定性量化专委会常务委员；科技部项目评估中心特邀评审专家。在国内外学术期刊已经发表三十余篇论文。获得部级科技进步二等奖一项。近年来主要从事复杂装备相关的不确定度量化理论与应用研究。

Digital Twin Brain: simulating and assimilating the human brain

张楠

Abstract：

The team led by Prof. Jianfeng Feng at Fudan simulated the human brain at the scale of up to 86 billion neurons with 100 trillion neural synapses: the digital twin brain (DTB), which mimics certain aspects of its biological counterpart both in the resting state and in action. A novel partition layout of neurons to 14,012 GPUs (57.39 million cores) and a two-level routing scheme between GPUs to implement simulations and a hierarchical mesoscale data assimilation method was developed capable of constructing more than trillions of parameters from the estimated hyper-parameters. The constructed DTB can reproduce blood-oxygen-level-dependent (BOLD) signals of its resting-state biological counterpart with a very high correlation (> 0.9). We enable the DTB to interact with its environment by demonstrating some potential applications in visual and auditory task. The result demonstrates the feasibility of implementing a digital twin of the human brain, overcoming key computational, mathematical and biological challenges and the broad range of applications of the DTB

个人简介：

张楠，青年副研究员。2015年获得美国德州A&M大学统计学博士学位并留校从事博士后研究。入选2017年度上海市青年科技英才扬帆计划。其研究成果已发表在Biometrika、Statistica Sinica等顶级期刊上。主要研究方向：非参数统计方法、函数型数据分析、基于抽样方法的大数据分析，以及在地球物理、生物信息、工程管理等领域的统计应用。

A Unified PCA Framework for Irregularly Observed Functional Time Series

黄辉

Abstract：

Irregularly observed functional time series (FTS) are increasingly available in many real-world applications. To analyze FTS, it's crucial to account for both serial dependencies and the irregularly observed nature of functional data. However, existing methods for FTS often rely on specific model assumptions in capturing serial dependencies, or cannot handle the irregular observational scheme of functional data.To solve these issues, one can perform dimension reduction on FTS via functional principal component analysis (FPCA) or dynamic FPCA. Nonetheless, these two methods may either be not theoretically optimal or too redundant to represent serially dependent functional data. In this article, we introduce a novel dimension reduction method for FTS based on the framework of dynamic FPCA. Through a new concept called optimal functional filters, we unify the theories of FPCA and dynamic FPCA, providing a parsimony and optimal representation for FTS adapting to its serial dependence structure. This framework is referred to as principal analysis via dependency-adaptivity (PADA). Under a hierarchical Bayesian model, we establish an estimation procedure for dimension reduction via PADA. Our method can be used for both sparsely and densely observed FTS, and is capable of predicting future functional data. We investigate the theoretical properties of PADA and demonstrate its effectiveness through extensive simulation studies. We illustrate our method via dimension reduction and prediction of daily PM2.5 data.

个人简介：

黄辉教授，研究方向时空数据分析，函数型数据分析，环境统计学，空间流行病，参与过多项国家级重大重点项目，发表过十多篇论文。黄辉博士2004年本科毕业于中国科学技术大学统计与金融系，2010年于美国马里兰大学巴尔地摩郡分校获统计学博士学位，随后分别在耶鲁大学公共卫生学院和迈阿密大学商学院从事博士后研究工作。2013年加盟北京大学统计科学中心、数学科学学院概率统计系，2017加入中山大学数学学院，担任教授、博士生导师，现为中国人民大学教授。现主要从事空间统计学，时空分析，函数型数据分析，图谱分析与空间点模式分析等领域的教学和研究工作，在空间流行病和大气污染的时空统计建模方面有大量研究经验。目前担任中国现场统计研究会空间统计分会理事、IBS-China青年理事。

油气渗流历史拟合与不确定性量化

张钊

摘要：

油藏数值模拟是不确定性量化的重要应用领域。由于储层物性参数的非均质性以及实际测量的稀疏性，造成储层物性参数具有很高的内在不确定性。油藏工程领域通常利用地质统计学方法构建一系列地质模型来反映这种不确定性。然而，基于这些先验地质模型进行油藏数值模拟得到的动态响应往往与实际生产观测数据不一致。对此，需对地质模型进行历史拟合，使得模拟结果能够匹配观测数据。这是典型的贝叶斯反问题不确定性量化。对此，难点一是油藏数值模拟通常计算量很大，而地质参数则往往维度较高，造成反演计算量非常大。难点二是很多地质参数的概率分布复杂甚至不确定，造成反演与数据同化算法设计困难。对此，我们进行了如下几方面工作。首先，在物理信息神经网络（PINN）基础上，提出了离散化PINN方法，克服了强非均质油藏上自动微分误差大的问题。大部分现有历史拟合方法是点估计，而MCMC计算量太大。为此，在变分贝叶斯方法基础上，结合数据同化提出了集合变分贝叶斯方法，克服了随机变分贝叶斯计算量太大以及不稳定的问题，实现了高效的油气渗流历史拟合与不确定性量化。进一步，针对参数的分布不确定性问题，借鉴非线性期望理论，提出了非线性期望推断方法，并实现了复杂多点统计学模型的历史拟合与不确定性量化。

个人简介：

张钊，山东大学数学与交叉科学研究中心副研究员、山东大学未来计划青年学者。2010年本科毕业于山东大学数学学院，2014年获得英国拉夫堡大学博士学位并于2014-2018年在英国赫瑞瓦特大学从事博士后研究。主要研究方向为不确定性量化、计算流体力学与机器学习，在Journal of Computational Physics等高水平期刊和会议发表论文10多篇。

Sequential Decomposition of Multiple Seasonal Components Using Spectrum-Regularized Periodic Gaussian Process

李勇祥

Abstract：

Many real-world time series, such as electricity demand data, biomedical signals, and mechanical vibration signals, exhibit complex trends, encompass multiple seasonal (or periodic) components, and are prone to noise contamination. Existing decomposition methods encounter difficulties when confronted with unknown periods and the presence of multiple nonlinear seasonal components. To address these challenges, we propose a novel nonparametric approach based on periodic Gaussian process models, called sequential seasonal-trend decomposition (SSTD). This model is capable of extracting multiple seasonal components sequentially while estimating the component periods. A spectrum-regularized periodic Gaussian process is proposed to sequentially extract each of the seasonal components, leveraging Fourier basis functions to represent the remaining components. The unknown periods are estimated through a tailored two-step parameter estimation technique from the non-convex likelihood. To mitigate the computational complexity of the proposed method, we propose a circulant acceleration approach. By enabling the sequential extraction of multiple seasonal components and the estimation of unknown periods, SSTD bridges a gap in existing methodologies, yielding improved accuracy and efficiency. Empirical studies on synthetic and real-world data demonstrate its outperformance over current methods.

个人简介：

李勇祥，上海交通大学工业工程与管理系，副教授。围绕复杂系统的质量与可靠性，主要研究试验设计与分析、统计与机器学习、统计质量控制、统计信号处理等数据科学方法与大数据技术。2021年入选上海市浦江A类人才计划。现任INFORMS Journal on Data Science副主编。

High-dimensional Ensemble Kalman Filter with Localization, Inflation and Iterative Updates

孙浩轩

Abstract：

Accurate estimation of forecast error covariance matrices is an essential step in data assimilation, which becomes a challenging task for high-dimensional data assimilation. The standard ensemble Kalman filter (EnKF) may diverge due to both the limited ensemble size and the model bias. In this paper, we propose to replace the sample covariance in the EnKF with a statistically consistent high-dimensional tapering covariance matrix estimator to counter the estimation problem under high dimensions. A high-dimensional EnKF scheme combining the covariance localization with the inflation method and the iterative update structure is developed. The proposed assimilation scheme is tested on the Lorenz-96 model with spatially correlated observation systems. The results demonstrate that the proposed method could improve the assimilation performance under multiple settings.

个人简介：

孙浩轩是北京大学前沿交叉学科研究院大数据研究中心的博士生，目前主要的研究方向是数据同化中的高维统计方法。

数字孪生技术在设备管理和工艺优化中的应用

徐徐

摘要：

智能制造是现代工业发展的必经之路，是保障水泥、钢铁等大型流程型企业实现高质量发展的前提和基础。对于大型生产企业来说，设备的正常高效运行至关重要。设备健康管理通过对人工智能算法与故障机理、工艺流程相结合的方法，研发生产设备故障智能预警、产线健康评估、机理诊断等模型，为实现精准维护提供决策支持，有效保证了产线安全高效运行。在实现设备稳定运行的基础上，通过对生产工艺中过程数据的高效采集和处理，采用自学习和自适应功能的智能预测控制算法，对系统关键变量进行优化，实现稳定生产、节能降耗、降低操作人员劳动强度，具有很高的经济和社会价值。本报告也将介绍相关实际应用案例。

个人简介：

徐徐，长三角信息智能创新研究院工业智能首席专家，南京凯奥思数据技术有限公司副总经理，国家重点人才工程A类青年人才。主要研究方向是数据分析建模方法在工业现场的应用，包括预测性运维、先进控制、库存优化、智能调度等。

探索数字孪生技术和贝叶斯优化在车载语音技术中的应用

陈代君

摘要：

随着智能汽车和车载语音技术的快速发展，数字孪生技术和贝叶斯优化方法正日益成为提升语音识别性能和用户体验的重要工具。本报告首先介绍了如何基于用户的车载语音日志数据和实时场景数据，利用数字孪生技术提升个性化推荐系统的性能，并为汽车制造商创建商业智能报告。同时，结合贝叶斯优化算法（如迁移贝叶斯优化和带约束的贝叶斯优化），对车载语音识别模型的多个参数进行调优，以提高识别率和响应速度。实验结果表明，在资源和时间有限的条件下，迁移贝叶斯优化能够迅速提升目标语音识别模型的性能。最后，本报告探讨了数字孪生技术在未来车载应用中的可能方向，包括个性化语音助手和个性化保养策略。同时，分析了贝叶斯优化在新一代端到端语音识别模型中的潜在应用前景。

个人简介：

陈代君是 Cerence Inc（Nuance 分拆的车载业务公司）语音识别团队的科学研究经理。2018年6月获得中国科学院数学与系统科学研究院的博士学位，随后加入 Nuance Inc，参与汽车问答系统和语音识别系统的开发与研究。陈代君的研究方向包括贝叶斯优化、计算机实验、语音识别和机器学习等。

Remaining Useful Life Prediction of Lithium-ion Batteries Using Monotone Decomposition

王典朋

Abstract：

Accurately predicting the remaining useful life (RUL) of lithium-ion batteries is vital for efficient equipment health management. Throughout the aging process, the battery capacity exhibits nonlinear behavior, with intermittent capacity regeneration phenomena causing sudden increments between consecutive cycles, posing challenges for modelling and prediction. Despite the frequent use of empirical mode decomposition (EMD) to decompose capacity series, most EMD-based RUL prediction methods encounter limitations including end effects and a lack of uncertainty quantification. To address these challenges, we introduce a novel RUL prediction framework, MonoD-GPR-DeepAR, featuring a unique data decomposition algorithm, MonoD. MonoD alleviates end effects by decoupling the original capacity signal into a smooth, decreasing trend and a fluctuant capacity regeneration term. Gaussian process regression (GPR) and deep autoregressive (DeepAR) models are then applied to the subseries for prediction, including uncertainty intervals. Validation using NASA lithium-ion battery degradation datasets demonstrates MonoD's superior performance in capturing the authentic aging trajectory characteristics. The MonoD-GPR-DeepAR model excels compared to alternative methods, showing its effectiveness in addressing complexities introduced by capacity regeneration phenomena in lithium-ion battery RUL prediction.

个人简介：

王典朋，北京理工大学数学与统计学院，副教授，博士生导师，统计系系主任。主要研究序贯试验设计、计算机试验设计和工业统计，获得中国数学会概率统计学会“钟家庆优秀论文奖”，现任北京大数据学会常务理事、现场统计研究会多元分析专业委员会常务理事。

Towards Fast Twinning: Surrogate Models and Engineering Applications

岳小伟

Abstract:

Fast Twinning is one core element for Digital Twins. It refers to the expedited and real-time pairing/updating of the virtual representation so that the digital twin simulator can reflect the updates in the physical counterpart efficiently. Surrogate models, or emulators, can tackle the high computational cost in high-fidelity digital twin, and have the potential to be a fundamental module for digital twins. We introduce some practical engineering examples of developing and implementing surrogate models and uncertainty quantification for finite element simulation, aerospace applications, and wind energy operations, etc.

个人简介：

Dr. Xiaowei Yue is an associate professor in the Department of Industrial Engineering at Tsinghua University. His research interests focus on engineering-driven data analytics for quality engineering. His research has obtained more than 10 best paper awards and two best dissertation awards. Dr. Yue is a recipient of multiple awards, such as IISE Hamed K. Eldin Outstanding Early Career IE in Academia Award, SME Outstanding Young Manufacturing Engineer award, IISE Manufacturing & Design Outstanding Young Investigator Award, and Grainger Frontiers of Engineering Grant Award from the U.S. National Academy of Engineering, etc. He serves as an associate editor for IISE Transactions, IEEE TASE, IEEE TNNLS, Journal of Intelligent Manufacturing, and ASME JCISE. He is also selected to be an editorial board member for PNAS Nexus, an open-access journal of the U.S. National Academy of Sciences (NAS). He is selected to be the President-Elect for the IISE Quality Control and Reliability Engineering Division in 2024.

Novel Bayesian optimization algorithms for robust product design with application to aerospace engineering

韩梅

Abstract:

In aerospace design activities, manufacturing and environmental variations are commonly encountered to deteriorate product quality. To make such aerospace systems less sensitive to such variations, robust engineering design methods, such as robust parameter design, are effective in selectinghigh-qualityproductdesign by considering input variations. With rapid developments in computers, simulation models have been commonly used in engineering design, and Bayesian optimization is popular in optimizing time-consuming computer simulations.In this presentation, we shall discuss developing efficient Bayesian optimization algorithms to actively learn of a contour for integrated responses, better minimize the expected quality loss in robust parameter design with parallel computing techniques, and more efficiently solve the multi-objective stochastic optimization problem in engineering design.Novelacquisition functions to choose values of both control and noise factors for determining the follow-up design points are proposed. Fast calculation formulas for the proposed acquisition functions are further developed to speed up optimization. Numerical examples and case studies in robust airfoil design are also presented to demonstrate the advantages of the proposed Bayesian optimization algorithms.

个人简介：

韩梅，博士，南京航空航天大学经济与管理学院副教授，硕士生导师。本科毕业于华中科技大学自动化专业，博士毕业于香港城市大学系统工程与工程管理专业。研究方向包括工业工程中的统计质量管理、计算机试验设计与优化、及质量管理和控制的应用等。担任中国优选法统筹法与经济数学研究会工业工程分会理事、 中国系统工程学会系统可靠性分会委员、中国运筹学会可靠性分会青年委员 ，国际期刊IISE, IJPR, IEEE ITS等学术期刊审稿人。主持国家自科基金项目2项（面上项目1项和青年项目1项），主持省部级及其他项目8项，研究成果发表在《Journal of Quality Technology》、《IISE Transactions》、《Swarm and Evolutionary Computation》等国际顶级与权威学术期刊。  

Local transfer learning Gaussian process modeling, with applications to surrogate modeling of expensive computer simulators

王新明

Abstract:

A critical bottleneck for scientific progress is the costly nature of computer simulations for complex systems. Surrogate models provide an appealing solution: such models are trained on simulator evaluations, then used to emulate and quantify uncertainty on the expensive simulator at unexplored inputs. In many applications, one often has available data on related systems. For example, in designing a new jet turbine, there may be existing studies on turbines with similar configurations or geometries. A key question is how to transfer information from such ``source'' systems for effective surrogate training on the ``target'' system of interest. We thus propose a new LOcal transfer Learning Gaussian Process (LOL-GP) model, which leverages a carefully-designed Gaussian process to transfer such information for surrogate modeling. The key novelty of the LOL-GP is a carefully-integrated regularization model, which identifies regions where transfer should be performed and regions where it should be avoided. This ``local transfer'' property is desirable for scientific systems: at certain parameters, such systems may behave similarly and thus transfer is beneficial; at other parameters, they may behave differently and thus transfer is detrimental. By accounting for local transfer, the LOL-GP may remedy an important limitation of ``negative transfer'' in existing transfer learning models, where the transfer of information worsens predictive performance. We derive a Gibbs sampling algorithm for efficient posterior predictive sampling on the LOL-GP. We then show, via a suite of numerical experiments and an application to jet turbine design, the improved surrogate performance of the LOL-GP over the existing state-of-the-art.

个人简介：

王新明，北京大学工业与系统工程专业博士生，研究方向为：迁移学习方法、贝叶斯模型及其在智能制造、数值仿真等领域的应用。曾获INFORMS年会QSR方向最佳论文奖、data mining方向最佳论文提名等荣誉。本次报告内容为：局部迁移学习的高斯过程代理模型，指导教师为来自Duke大学的Prof. Simon Mak。

Efficient calibration of computer models with multivariate output

孙杨

Abstract:

The classical calibration procedures of computer models only concern the univariate output, which would not be satisfied in practice. Multivariate output is gradually more prevalent in a wide range of real-world applications, which motivates us to develop a new calibration procedure to extend the classical calibration methods to multivariate cases. In this work, we propose an efficient calibration procedure for multivariate output within restricted correlation. First, we construct an estimator of the discrepancy function between the true process and the computer model by the local linear approximation, then obtain an estimator of the calibration parameter by the weighted profile least squares and establish its asymptotic properties. In addition, we also develop an estimator of the calibration parameter in a special situation, whose asymptotic normality has been derived. Numerical studies including simulations and an application to composite fuselage simulation verify the efficiency of the proposed calibration procedure.

个人简介：

孙杨，2024年7月博士毕业于北京大学数学科学学院统计学专业，师从房祥忠教授；现为北京大学数学科学学院博士后，合作导师为陈松蹊院士，研究兴趣为不确定性量化、时空统计、海洋大数据分析等。已在Journal of Multivariate analysis, Journal of

System Science and Complexity, Statistical Papers等期刊上发表论文7篇。

针对超材料函数型响应的建模

季春霖

摘要：

在超材料的结构设计中，我们对给定的设计点进行有限元分析和模拟，通常所得到的响应是复杂的函数曲线：随频率变化的多谐振曲线。传统的多项式或样条代替模型不适合处理这种谐振曲线的拟合。我们提出了一些半物理模型，将曲线分解为一些有“物理”意义的项，获得了高精度的曲线拟合。并且得到了一种对这类复杂曲线的有效的回归分析方法。此外我们讨论了在用贝叶斯优化超材料时，如何处理这类函数型响应，给出了局部响应匹配和全局响应匹配两种目标函数设计的方法。深度学习方法的发展为复杂曲线拟合和材料设计带来了新的机遇，我们探索了利用条件变分推断网络实现由函数曲线直接预测微结构拓扑，已经取得初步成效，为函数型数据分析和处理带来了新的方法。仿真研究表明了上述方法的有效性。

个人简介：

季春霖，美国杜克大学统计科学博士，中国国籍，无境外永久居留权。2004年3月至2004年12月，任中国香港中文大学电子工程系助理研究员；2006年4月至2006年8月，任中国香港城市大学电子工程系助理研究员；2010年1月至2010年2月，任美国哈佛大学统计系博士后研究员；2010年3月至今，任光启理工研究院副院长、核心科学家，现任光启技术股份有限公司董事。

Baysian optimization in materials science with emphasis on adaptive sampling using uncertainties for targeted design

薛德祯

Abstract：

One of the primary challenges in materials discovery is efficiently navigating the vast search space to identify materials with targeted properties, as trial-and-error approaches are impractical. We review how information science methods can accelerate the search and discovery of new materials. Specifically, Bayesian optimization enables us to effectively explore the search space iteratively, identifying promising candidates to guide experiments and computations. This approach leverages uncertainties and predictions from a surrogate model, combined with a utility function that prioritizes decision-making for unexplored data. We discuss several utility functions and demonstrate their application in materials science, impacting both experimental and computational research. Finally, we summarize by addressing generalizations to multiple properties and multifidelity data, and identify challenges, future directions, and opportunities in the emerging field of materials informatics.

个人简介：

薛德祯，西安交通大学教授、博导，2013年1月至今在金属材料强度国家重点实验室工作。主要从事材料信息学领域研究，致力于材料学与信息学的交叉融合，在机器学习辅助实现智能材料高性能化方面展开了系统深入的研究，具有扎实的理论基础与丰富的研究经验。发表论文 120 余篇，其中以第一/通讯作者发表在 Nat. Comm.、Phys Rev. Lett.、PNAS.、Adv. Mater.等期刊发表60 余篇（5篇 ESI 高被引论文）；担任《Materials Genome Engineering Advances》、《Advanced Engineering Materials》、《Journal of Materials Informatics》《Crystals》青年编委/编委，材料基因工程国际高层论坛学术委员会副主任。入选了国家层次青年人才计划，获得了教育部自然科学奖一等奖（排名第六）、陕西省优秀博士论文、美国洛斯阿拉莫斯国家实验室院长博士后，陕西省高校青年人才支持计划等荣誉。

The Adaptive Surrogate Models Based onDynamic

 Gaussian Process Regression

段晓君

Abstract：

Modeling and analyzing uncertainty in complex systems is essential for understanding system features. However, experiment-based quantification of this uncertainty faces challenges due to the high cost and limited experimental duration, restricting the amount of data collected. To address these limitations, we propose several adaptive surrogate modeling approaches using dynamic Gaussian process regression. Our methodology includes the Heuristic-based Annealing Combinable Gaussian Process (ACGP), which dynamically selects the model structure for complex parametric systems, improving modeling precision and identifying key parameters. Additionally, the Gradient-enhanced Sequential Surrogate Modeling (GRASE-SUMO) dynamically reduces the parameter space to model spatially unstable systems, enhancing accuracy and computational efficiency. Finally, the Partitioned Spatio-Temporal Gaussian Process Regression (PSTGPR) dynamically discovers local spaces to model spatio-temporally unstable systems, capturing system change points and coupling structures with high accuracy at a lower cost. These data-driven methods enable the construction of adaptive surrogates that accurately characterize complex systems, facilitating a deeper understanding of their performance under various conditions.

个人简介：

段晓君，国防科技大学理学院，教授。主要研究方向为复杂系统建模，试验设计与不确定性量化。入选教育部新世纪优秀人才支持计划，省部级高层次创新人才工程。中国现场统计研究会试验设计分会常务理事，中国工业与应用数学学会理事，中国数学会均匀设计分会常务理事。

交通仿真的基础问题与近期发展

魏贺

摘要：

仿真在交通管理、出行优化等方向具有非常重要的作用，对于智能交通发展、数字孪生建设等有非常大的贡献。该报告对交通仿真的发展概况进行简要介绍，随后对比国内外交通仿真的理论发展，从健全规则、现实挑战等方面进行阐述，最后指出要持续加强学科间、理论与实践间的互动联系与交叉融合，建设“数学+”综合计算与仿真平台逐步实现弯道超车。

个人简介：

魏贺，就职于北京市城市规划设计研究院交通规划所，担任主任工程师，研究方向为交通战略与政策、交通模型与大数据应用、国际前沿观察。获得省部级以上奖励10余项、授权发明专利2项、参与专业书籍编著6部、独立及合作发表学术论文近40篇，开设“真武草堂关门弟子”专业公众号。兼任中国城市交通规划专委会委员、中国交通建模与仿真专委会委员、科技核心期刊《城市交通》编委及审稿人、《交通与港航》期刊审稿人、北京建筑大学硕士生指导导师。

不完全观测下的个体轨迹重构：基于自适应期望最大化注意力算法的时空概率图模型

孙璇

摘要：

城市轨道交通自动车辆定位（Automatic Vehicle Location，AVL）数据轨迹时空信息对运营策略调整、个性化推荐和应急指挥决策具有重要意义。然而，由于缺乏行程观测，仅基于自动售检票（Automatic Fare Collection，AFC）数据和自动车辆定位（Automatic Vehicle Location，AVL）数据很难准确地从轨迹推断出未知信息。为了解决这一问题，本文提出了一种基于自适应期望最大化注意力的时空概率图模型(STPGM-AEMA)实现不完全观测信息下的个体轨迹重构问题。该方法包括三个步骤：首先，通过数据挖掘和组合枚举得到乘客的潜在的具有一一映射关系的潜在的可登乘列车候选解和走行时长候选解；其次，为刻画个体出站走行时长状态值和系统出站走行时长分布两者的交互影响效应，引入表征个体潜在候选解的全局隐变量和表征系统分布特征的局部隐变量模拟所有乘客个体状态与系统状态的时空依赖关系，构建个体车次推断的时空概率图模型（STPGM）；然后，考虑乘客走行时长状态值的非随机缺失，提出了一种注意机制增强的期望最大化算法（AEMA）实现个体轨迹中出站走行时长状态值和上车车次的推断，并结合马尔科夫链蒙特卡洛（MCMC）的随机仿真策略完成个体轨迹重构过程。最后，利用典型的OD对数据集间AFC、AVL和实际个体轨迹追踪数据验证了所提方法的有效性。结果表明，在小样本训练场景下，本文所提STPGM-AEMA方法对个体轨迹推断精度提高15%，一致性提升20%。

个人简介：

孙璇，女，北京交通大学博士生，在先进轨道交通自主运行全国重点实验室贾利民、秦勇教授主动安全保障团队从事出行状态监测与仿真、风险辨识与应急决策模拟方向研究。任国际期刊Smart and Resilient Transportation主编助理，ITS智能轨道交通发展专业委员会学术秘书。参与国家及省部级等项目10余项，主持校级研究生创新项目1项。录用及发表学术论文14篇（SCI 6篇），公开及授权发明专利18项（排名前三6项），授权软件著作权5项（排名前三2项），参与编撰书籍《计算智能与交通运输工程应用》、《城市轨道交通路网客流预测仿真与协同组织优化方法》。曾获第三届北京市应急管理领域优秀博士生（1/10人），第三届中国城市轨道交通创新创业大赛优秀项目奖（排名1），北京市“挑战杯”专项赛铜奖（排名1），第四届北马其顿国际专利发明金奖，多次获“优秀党员”、“三好学生”等荣誉。

地理加权回归模型的推广、模型推断及其在交通中的应用

吴颖

摘要：

在空间统计中，地理加权回归模型能够用来分析数据的空间效应，在地理、交通、经济等领域有着广泛的应用。首先，我们提出了一类广义地理加权回归（UGWR）模型，使其满足复杂空间数据的分析需求。UGWR模型同时考虑了空间相关性和空间异质性，允许响应变量和扰动项具有空间自回归性，且协变量系数和空间自回归系数能够随空间变化，我们给出了基于重构参数化方法的模型估计、预测方法。其次，针对变系数地理加权回归模型，我们提出了三种贝叶斯推断方法，包括传统高斯过程先验、重构参数化（RP）先验及其简化版本。数值实验表明，简化的方法能够在更低的计算成本下获得较好的推断结果，适合处理大规模数据集。基于上述方法，我们分析了北京出租车乘载量数据集，探究不同时段不同天气下的城市出租车需求变化，为城市出租车调度提供了参考。

个人简介：

吴颖博士于2019年在山东大学数学学院获得理学学士学位，于今年六月份在中国科学院数学与系统科学研究院获得博士学位，现加入香港中文大学(深圳)吴建福院士团队从事博士后研究工作。她的主要研究方向是计算机实验与不确定性量化，空间统计，贝叶斯分析等。

BdryGPs: incorporating general boundary information via SPDE approach

丁亮

Abstract：

Gaussian Processes (GPs) are widely used as surrogate models for emulating computer codes that simulate complex physical phenomena. In many problems, additional boundary information (i.e., the behavior of the phenomena along input boundaries) is known beforehand, either from governing physics or scientific knowledge. In this work, we propose a framework of GP models based on stochastic Partial Differential Equations (SPDEs) with Dirichlet boundary conditions. For general SPDEs, we show that their covariances can be represented in a path integral form. Based on this path integral form, we provide a regression approximation, called regression kernel, along with the necessary criteria to accurately approximate the SPDE covariances. We also show that for covariances in tensor form, closed-form solutions are available. Numerical experiments are conducted to illustrate that incorporating boundary information can significantly improve performance of the model.

个人简介：

丁亮，复旦大学大数据学院青年研究员。2019年获得香港科技大学运筹学博士学位，2018-2019为佐治亚理工学院工业与系统工程系访问学者。之后在美国德州A&M大学从事博士后研究。其主要成果发表于Operations Research, JMLR等国际期刊。

Sobolev Calibration of Imperfect Computer Models

王文佳

Abstract:

Calibration refers to the statistical estimation of parameters in computer experiments, which is often intrinsic and hard to measure with physical tools. This work develops a novel calibration method for imperfect computer models, called Sobolev calibration. The method allows great flexibility for practitioners to achieve both “point-wise value” and “overall shape” approximation, and leads to better physical experiments reconstruction. We prove that the Sobolev calibration enjoys desired theoretical properties including fast convergence rate, asymptotic normality and semiparametric efficiency. We also show that the Sobolev calibration serves as a bridge between two prominent methods: $L_2$ calibration and Kennedy and O'Hagan's calibration method. Additionally, we theoretically justify that our method is applicable to the case when the physical process is indeed a Gaussian process, which follows the original idea of Kennedy and O'Hagan's. Numerical simulations as well as a real-world example illustrate the performance of the proposed method.

个人简介：

王文佳是香港科技大学（广州）信息枢纽数据科学与分析学域的助理教授；2018年8月获得佐治亚理工学院工业工程系博士学位。王文佳的研究方向包括不确定性量化、计算机实验、机器学习、随机仿真和非参数统计。目前已在统计学、机器学习、管理学顶级期刊、会议Journal of the American Statistical Association，Journal of Machine Learning Research，Management Science，Technometrics，StatisticaSinica，NeurIPS，ICLR，ICML等发表多篇文章。

Large-Scale Low-Rank Gaussian Process Prediction with Support Points

代文林

Abstract:

Low-rank approximation is a popular strategy to tackle the “big n problem” associated with large-scale Gaussian process regressions. Basis functions for developing low-rank structures are crucial and should be carefully specified. Predictive processes simplify the problem by inducing basis functions with a covariance function and a set of knots. The existing literature suggests certain practical implementations of knot selection and covariance estimation; however, theoretical foundations explaining the influence of these two factors on predictive processes are lacking. In this paper, the asymptotic prediction performance of the predictive process and Gaussian process predictions is derived and the impacts of the selected knots and estimated covariance are studied. We suggest the use of support points as knots, which best represent data locations. Extensive simulation studies demonstrate the superiority of support points and verify our theoretical results. Real data of precipitation and ozone are used as examples, and the efficiency of our method over other widely used low-rank approximation methods is verified.

个人简介：

代文林博士2018年加入中国人民大学统计与大数据研究院，现为副教授、博士生导师。此前，他曾在沙特阿卜杜拉国王科技大学(KAUST)进行博士后研究，2014年于香港浸会大学获得统计学博士学位，2008年于北京理工大学获得统计学学士学位。研究兴趣包括非参数统计、函数型数据探索性分析、网络数据与时空数据分析及统计模型在环境、工业、经济等领域的应用。

Asymptotic Theory for Linear Functionals of Kernel Ridge Regression

邹璐

Abstract:

An asymptotic theory is established for linear functionals of the predictive function given by kernel ridge regression, when the reproducing kernel Hilbert space is equivalent to a Sobolev space. The theory covers a wide variety of linear functionals, including point evaluations, evaluation of derivatives, $L_2$inner products, etc. We establish the upper and lower bounds of the estimates and their asymptotic normality. It is shown that $\lambda\sim n^{-1}$ is the universal optimal order of magnitude for the smoothing parameter to balance the variance and the worst-case bias. The theory also implies that the optimal $L_\infty$ error of kernel ridge regression can be attained under the optimal smoothing parameter $\lambda\sim n^{-1}\log n$. These optimal rates for the smoothing parameter differ from the known optimal rate $\lambda\sim n^{-\frac{2m}{2m+d}}$ that minimizes the $L_2$ error of the kernel ridge regression.

个人简介：

邹璐，香港科技大学研究助理。2019年获得香港科技大学运筹学博士学位。2018-2021为美国德州A&M大学工业与系统工程系访问学者。

Deep digital twins based on GPtime-GAN for prognostics and health monitoring

石坚

摘要：

对time-GAN进行了生成效果上改进，并通过改进后的网络GPtime-GAN对一些工业资产构造深度数字孪生体。另一方面，利用深度数字孪生内部结构建立了具有概率意义的健康指标，并给出了健康指标的置信下限。实例分析的结果表明，新的健康指标同时结合置信下限，可以使故障预警更加精确和稳定。

个人简介：

石坚，中国科学院数学与系统科学研究院研究员，中国科学院大学岗位教授，博士生导师。目前任中国现场统计研究会副理事长、中国概率统计学会常务理事。主要研究方向包括：统计推断、生物医学统计、工业统计、大数据统计和体育统计。已发表学术论文70余篇，包括JASA和Biometrika等国际知名统计学刊物，出版英文专著1本。主持和参加过国家自然科学基金面上与重点项目，以及科技部重大专项等应用课题。

基于虚拟维修的装备维修性设计与评估技术

伊枭剑

摘要：

以某装备动力舱为例，以提升动力舱拆装作业效率为目标，以维修作业规划和人机工效分析为理论基础，通过拆装过程的分析、规划和可视化，指导作业安排，规范作业姿态与作业方法，并对作业量进行科学估计，从而支撑动力舱设计初期开展维修性设计与验证工作，设计后期实现“虚拟人拆装虚拟动力舱”的作业过程仿真，为动力舱的维修性与维修设计工作提供关键的技术支持。

个人简介：

伊枭剑，北京理工大学长聘副教授，硕士生/博士生导师，研究方向：复杂系统可靠性与智能决策、故障预测与健康管理技术，至今在该领域已发表SCI期刊学术论文40余篇，多项科技成果被鉴定为国际先进水平，并获得省部级科技进步二等奖2项、三等奖1项。在基础科研方面，作为项目负责人主持国家自然科学基金、国防科工局技术基础、国防科工局基础科研、装备预先研究共用技术和装备技术基础科研项目等10余项。在工程应用方面，在多个型号中担任通用质量特性专项组组长、总体组副组长、主任设计师和副主任设计师等职务；目前，作为项目负责人主持某轮式装甲车健康管理系统研制、某履带式战车状态监测与评估系统研制等；并作为主任设计师承担国家XX工程任务2项。在学术推广方面，目前为中国技术市场协会工程故障预测与健康管理技术专业委员会理事、中国现场统计研究会可靠性工程分会理事，中南大学学报青年编委，Reliability Engineering and System Safety、Quality and Reliability Engineering International等10余个SCI期刊审稿专家，以及ELSEVIER能源可靠性专著的评审专家，担任国际会议SDPC & CLQRC—Beijing 2019大会主席、并以程序委员会主席服务于PHM领域三大国际会议。

Parameter Inference via Nonlinear Partial Differential Equations Informed Gaussian Processes

李赵辉

Abstract:

Partial differential equations (PDEs) are widely used for description of physical and engineering phenomena. Some key parameters involved in PDEs, which represents certain physical properties with important scientific interpretations, are difficult or even impossible to be measured directly. Estimation of these parameters from noisy and sparse experimental data of related physical quantities is an important task. Many methods for PDE parameter inference involve a large number of evaluations of numerical solution of PDE through algorithms such as finite element method, which can be time-consuming especially for nonlinear PDEs. We propose a novel method for estimating unknown parameters in PDEs, called PDE-Informed Gaussian Process Inference (PIGPI). Through modeling the PDE solution as a Gaussian process (GP), we derive the manifold constraints induced by the (linear) PDE structure such that under the constraints, the GP satisfies the PDE. For nonlinear PDEs, we propose an augmentation method that transfers the nonlinear PDE into an equivalent PDE system linear in all derivatives that our PIGPI can handle. PIGPI can be applied to multi-dimensional PDE systems and PDE systems with unobserved components. The method completely bypasses the numerical solver for PDE, thus achieving drastic savings in computation time, especially for nonlinear PDEs. Moreover, the PIGPI method can give the uncertainty quantification for both the unknown parameters and the PDE solution.

个人简介：

李赵辉，助理研究员，中国科学院数学与系统科学研究院。2014年在中国科学技术大学获得理学学士学位，2020年获得中国科学院数学与系统科学研究院和香港城市大学联合培养博士学位。后在香港城市大学和佐治亚理工学院从事博士后研究工作。2023年7月加入中国科学院数学与系统科学研究院担任助理研究员。他的主要研究领域为系统可靠性，计算机实验与不确定性量化。在IISE ransactions， Technometrics， SIAM Journal of Uncertainty Quantification，NeurIPS等重要期刊会议上发表过多篇论文。

Student Poster:

Poster Session 1 (8月26日，周一，13：00 — 14：00)

Poster Session 2 (8月27日，周二，13：00 — 14：00)

Poster Session 3 (8月29日，周四，13：00 — 14：00)

参会人员