学术报告

学术活动

学术报告
03/01 2023 Seminar
  • Title题目 EPR-Net: Constructing non-equilibrium potential landscape via a variational force projection formulation
  • Speaker报告人 李铁军(北京大学数学科学学院)
  • Date日期 2023年3月1日 10:30
  • Venue地点 南楼6620会议室
  • Abstract摘要
     In this talk, we present a novel yet simple deep learning approach, dubbed EPR-Net, for constructing the potential landscape of high-dimensional non-equilibrium steady state (NESS) systems. The key idea of our approach is to utilize the fact that the negative potential gradient is the orthogonal projection of the driving force with respect to an inner product weighted by the steady-state distribution. The minimum of the proposed loss function coincides with the steady entropy production rate (EPR) formula in NESS theory. The proposed learning strategy is also extended to dealing with dimensionality reduction and state-dependent diffusion coefficients in a unified fashion. It is succesfully applied to high-dimensional biochemical models with multi-stability, limit cycle, or strange attractor with non-vanishing noise. This is a joint work with Wei Zhang (ZIB) and Yue Zhao (PKU).

    个人简介:铁军,北京大学数学科学学院教授,国家自然科学基金委优秀青年及杰出青年基金获得者,入选教育部“新世纪优秀人才”支持计划。研究领域为随机模型及算法。在化学反应随机动力学算法、稀有事件及能量景观分析;单细胞转录组数据分析;复杂网络的社区结构等方面做出了重要成果。在包括PNAS, Nature Comm, PRX, CMP, JCP, SIAM系列等学术期刊共发表论文六十余篇。

    邀请人:周海军 研究员

附件下载: