Automated discovery of fundamental variables hidden in experimental data

发布时间：2023-09-15 作者：浏览次数：

Speaker:	黄旷	DateTime:	2023年9月18日（周一）上午10：00-12:00
Brief Introduction to Speaker: 研究领域为非局部模型的分析和计算;交通流的数学建模;平均野外游戏;用于动态系统建模的数据驱动方法.部分出版物：自动驾驶汽车速度控制的博弈论框架：桥接微观微分博弈和宏观平均场博弈（与X. Di，Q. Du和X. Chen），DCDS-B，2020。互联车辆非本地交通流模型的稳定性（与 Q. Du 合作），SIAM J. 应用数学，2022 年。自动发现隐藏在实验数据中的基本变量（与B. Chen，S. Raghupathi，I. Chandratreya，Q. Du和H. Lipson），Nat. Comput。科学， 2022.		Place:	6号楼2楼报告厅
Abstract:Physical laws can be described as relationships between state variables that give a complete and non-redundant description of the relevant dynamical systems. Most data-driven methods for modeling physical phenomena assume that observed data streams already correspond to given state variables. However, despite the prevalence of computing power and AI, the process of identifying a set of state variables themselves from experiment data has resisted automation. We propose a framework for determining how many state variables an observed system is likely to have, and what these variables might be, directly from video streams. We also demonstrate the effectiveness of this approach using video recordings of a variety of dynamical systems, ranging from elastic double pendulum to fire flames.

上一条：后量子公钥密码-基于编码的密码体制
下一条： Singularity formation for the Landau-Lifshitz-Gilbert equation in dimension two