学术活动

Machine learning of small data

来源:力学与工程学院  作者:马晓梅     日期:2020/10/13 19:06:49   点击数:1128  

时间:2020-10-20 10:00

地点:牵引动力国家重点实验室红楼218

报告人:张统一

 

个人简介:

张统一,中国科学院院士,香港工程科学院院士,材料科学与固体力学专家,我国材料基因组工程、材料信息学和力学信息学的推动者,上海大学材料基因组工程研究院创院院长,中国材料学会材料基因组工程分会首任主任。曾任香港科技大学讲座教授、方氏冠名教授,国际断裂学会副主席,远东及大洋洲断裂学会副主席。获香港裘槎高级研究学者奖、美国ASM International Fellow奖、国际断裂学会Fellow奖、国家自然科学二等奖(两次)、中国科学技术协会青年科技奖、何梁何利基金科技进步奖。近年来在国际和国内大力推动材料信息学并首提力学信息学新概念;倡导融合专家知识的数据驱动新材料发现,材料正向设计和逆向设计相结合的新理念;呼吁发展以数据为中枢,向上支撑新材料研发和创新,向下加快产业制造生产智能化和信息化的新模式。

讲座内容:

The ceaseless and infinite development of nature science is an endless and continuous process during which mankind observes and summarizes the nature behaviors, and then gains knowledge. From data to knowledge is a quantum jump, during which key factors are extracted and classified into input features and output responses, and relationships between input features and output responses are analyzed and, if possible, explicitly expressed in mathematic equations. Materials data, especially the materials data of mechanical behaviors such as creep, fatigue, fracture etc., are often small and high dimensional. Domain knowledge plays a crucial role in machine learning of small data. An example is given in the presentation to emphasize the role of domain knowledge in machine learning of small data. The example analyzes the data of size-dependent strength of concrete. On the other hand, statistical leaning of small data might be able to evaluate theoretical models developed from domain knowledge.


热点新闻

交大视频

XML 地图 | Sitemap 地图