科学计算前沿，邀你共同探讨！AI for Science交流会来了！

欢迎从事计算机/人工智能或各领域科学计算方向的研究人员踊跃报名本次线下交流会我们邀请到国内外优秀的科研人员分享他们在相关领域的知识和研究成果，为大家提供一个学习和深入探讨的机会。

1. 飞桨AI for Science领域探索与产品建设现状

主讲人：张艳博，飞桨高级技术产品经理 

报告简介：介绍飞桨AI for Science整体产品建设情况

2. 飞桨PaddleScience V1.0功能介绍及使用指南

主讲人：何森森，飞桨工程师

报告简介：介绍飞桨PaddleScience V1.0版本功能及使用指南

1. 科学计算：融合多保真数据的复合神经网络

主讲人：孟旭辉，2017年博士毕业于华中科技大学能源与动力工程学院；2018年-2022年美国布朗大学应用数学系从事博士后研究工作, 合作导师为美国工程院院士George Em Karniadakis教授；2022年3月至今任华中科技大学数学与统计学院数学与应用学科交叉创新研究院副教授。主要研究方向为科学计算中的深度学习方法。截至目前已在JCP、CMAME、SIAM Review等期刊发表SCI论文20余篇，谷歌学术总引用2600余次，4篇论文入选ESI高被引论文，1篇论文为热点论文；担任JCP、SISC、CMAME、Nat. Comput. Sci .等期刊审稿人。

报告简介：The recent developments in machine learninghave also influenced the computational modeling of physical systems, e.g. ingeosciences and engineering. Generally, large numbers of high-fidelity datasets are required for optimization of complex physical systems, which may leadto computationally prohibitive costs. On the other hand, inadequatehigh-fidelity data result in inaccurate approximations and possibly erroneousdesigns. Multi-fidelity modeling has been shown to be both efficient andeffective in achieving high accuracy in diverse applications by leveraging boththe low- and high-fidelity data. In this talk, I will introduce a newlydeveloped deep learning algorithm for multi-fidelity data fusion as well astheir applications to function approximation and inverse PDE problems.

2. 世界模型：直觉物理推理与决策

主讲人：王韫博，上海交通大学助理教授，清华大学博士，CCF优博，国家自然科学基金原创探索项目负责人。主要从事机器学习与计算机视觉的研究，专注于可微物理模拟、有模型强化学习等方向，在TPAMI、NeurIPS、ICML、CVPR等CCF-A类期刊和会议上发表论文20余篇，2020年入选上海市青年科技扬帆计划。

报告简介：世界模型，是指智能体通过与环境交互，或通过观察物理世界的视觉表观演变，探索和推断其内部动力学模型，构建物理世界的可微分模拟器。近十年间，J. Tenenbaum、J. Schmidhuber、Y. LeCun等人分别从贝叶斯认知理论、视觉环境决策、自监督学习的角度给出了世界模型的不同表达形式。然而，如何在复杂的视觉场景中建立有效的世界模型仍然是一项开放课题。本次报告围绕世界模型的“观察”与“交互”两个方面，分别介绍“3D逆图形学物理推断”和“基于世界模型的强化学习”方法，相关工作分别发表于2022年的ICML和NeurIPS（Spotlight）。前者为流体动力学模拟提供了一种新思路，后者应用于DeepMind Control机械臂视觉控制任务和CARLA自动驾驶任务，为当前最佳基线模型之一。

主讲人：Madeleine Martinsen, Head of R&D Hoist & Underground Mining at ABB Service. More than 34 years of wide competences from engineering, internal auditing, controlling to management positions within both the Power and Automation Industry at ABB. Including abroad assignments in Germany, Ecuador, USA and Denmark. A trouble-shooter with successful experience from “turning around” businesses. As a person enthusiastic, she is open minded, optimistic, curios, analytic, structured and result oriented. Her driving force is best used within an organisation where daily operational excellence is at focus as for serving customers. Currently focusing on finalizing her PhD with the title ‘Monitoring of airflow and airborne particles, to provide early warning of irrespirable atmospherics conditions’.

主讲人：Erik Dahlquist (瑞典皇家工程院院士), Experienced Professor with a demonstrated history of working in the higher education industry. Skilled in Computer Science, Smart Grid, Biomass, Modeling, and Research and Development (R&D). Strong education professional with a PhD focused in Chemical Engineering from Kungliga tekniska högskolan.

报告简介：The industries production is predicted to completely develop into an autonomous operation in the future. Developing an autonomous manufacturing system can for some industries prove difficult due to inadequate connectivity and non-uniform manufacturing environment. This is the case for the mining business. To resolve the connectivity issue, the industries are investing in information system (IS) and information technology (IT), infrastructure to provide connectivity to the overall control system, which will support their digitalization journey. Anyhow, the transition to a fully autonomous mining will not happen over a night. A complete autonomous mining operation will demand solutions that can be trustworthy. Timely detection of faults, which can lead to harmful and dangerous situations, is of the most importance in an autonomous mining operation. Preventing and predicting the need for maintenance will play an important ingredient succeeding.