题目：结构工程中的机器学习：从数据驱动分析的巡游到新知识发现的可解释性和因果联系

Machine Learning in Structural Engineering: From Navigating the Realms of Data-driven Analysis to Explainability and Causal Knowledge Discovery

时间：2022年4月1日周五9：00－10：00

地点：腾讯会议 ID：93382698531

腾讯直播间：https://meeting.tencent.com/l/hA1SrLFrwFwl

报告人：M.Z.Naser（美国克莱姆森大学，土木与环境工程与地球科学学院）

M.Z.Naser助理教授官方主页：https://www.mznaser.com

欢迎广大师生参加！

土木与交通学院

2022年3月29日

报告人简介：

M.Z. Naser是美国克莱姆森大学土木与环境工程和地球科学学院的预聘助理教授，也是该校人工智能科学与工程研究所(AIRISE)的成员，和注册结构工程师。在加入克莱姆森大学之前，他在密歇根州立大学完成了博士学位，师从加拿大工程院院士Venkatesh Kudor教授。Naser教授的主要研究领域是结构工程和机器学习。至今已发表了80余篇期刊论文，包括四本书--由美国混凝土协会出版的《人工智能时代的混凝土行业》、由麦格劳-希尔出版的《结构防火工程》、泰勒和弗朗西斯出版的《利用人工智能进行工程、管理和基础设施安全》，以及施普林格出版出版的《火灾复原基础设施的认知和自主系统手册》。

 M.Z. Naser, assistant professor at the School of Civil and Environmental Engineering and Earth Sciences at Clemson University, a faculty member of the AI Research Institute for Science and Engineering (AIRISE), and a professional engineer. Prior to Clemson, he completed his Ph.D., at Michigan State University. Dr. Naser's primary research areas are structural engineering and causal machine learning. He has co-authored over 80 peer-reviewed publications, including four books – The Concrete Industry in the Era of Artificial Intelligence (Published by The American Concrete Institute), Structural Fire Engineering (published by McGraw Hill ), and Leveraging Artificial Intelligence in Engineering, Management, and Safety of Infrastructure (to be published by Taylor & Francis in 2022), and the Handbook of Cognitive and Autonomous Systems for Fire Resilient Infrastructures (to be published by Springer in 2022).

 报告摘要：

机器学习(ML)为复杂和独特的结构工程问题提供了新颖和令人兴奋的解决方案，其性能远远超过了传统方法和常规代码程序。然而，在土木工程培养计划和课程中却很少涉及机器学习。这就导致了，土木工程研究者和从业者只能作为机器学习的末端用户(即应用者)，无法完全掌握机器学习预测背后的推理逻辑，亦或是掌握如何剖析由黑箱衍生的决策，很难负责任地应用机器学习到实际问题中，甚至和涉及到值得我们关注和研究的工程科学和伦理问题。本次讲座将首先简单介绍下机器学习背后的基本概念，然后展示机器学习在结构工程的现状。之后，我们将介绍机器学习中的一些最新概念，包括数据驱动的ML、可解释的ML、绿色的ML、道德的ML和具有责任性的ML。最后，我们将分享ML作为知识发现系统的哲学观点，通过整合因果关系和因果发现的原则来推进土木工程领域发展。

Machine Learning (ML) generates novel and exciting solutions to complex and unique structural engineering problems with astonishing performance that far exceeds that of traditional methods and codal procedures. Yet, ML is rarely taught in a civil engineering curriculum. As such, we continue to primarily be secondhand ML users (i.e., appliers) and may not fully grasp the reasoning behind ML’s predictions, how to dissect its Blackbox-derived decisions, or how to responsibly apply ML. The above also brings scientific and ethical questions that warrant our attention, as well as investigation. This seminar hopes to chart a path that starts with a brief look into the big ideas behind ML and then showcase the state of ML in our domain. After that, we visit some of the latest concepts within ML in terms of data-driven ML, Explainable ML, Green ML, Ethical ML, and Responsible ML. Finally, we will share a philosophical view of ML as a Knowledge Discovery system that can help advance the domain of civil engineering via integrating principles of causality and causal discovery.