报告人：陈小丽（中国地质大学）

报告时间：2026年10月9日（星期五）16:30-18:30

报告地点：科技楼南楼706室

报告摘要：In the study of stochastic dynamical systems, quantifying noise-driven escape phenomena and transition paths is critical for understanding transitions between metastable states. Key metrics such as the mean first exit time, escape probability, and most probable transition path provide essential insights into these processes. However, traditional numerical methods, including Monte Carlo simulations and finite difference methods, suffer from low computational efficiency and poor generalizability, especially when dealing with systems with continuous parameter variations. Additionally, existing deep learning frameworks such as physics-informed neural networks require retraining as parameters change, severely restricting their performance in complex dynamic parameter systems. To address these challenges, this talk proposes a novel solution framework based on the deep operator network. It aims to efficiently solve escape problems and transition paths in stochastic systems across continuous parameter spaces. By parameterizing both the drift term and the noise intensity in the stochastic differential equation, this framework constructs an explicit mapping from the parameter space to the solution operator, eliminating the need for retraining when parameters change. Furthermore, this talk innovatively integrates reflective boundary conditions into the escape analysis of stochastic gene regulatory systems, offering a completely new analytical paradigm for investigating gene transcription dynamics.

报告人简介：陈小丽，中国地质大学（武汉）特任教授。2020年博士毕业于华中科技大学。2018年9月至2020年8月在美国布朗大学进行联合培养。2021年3月至2024年3月在新加坡国立大学从事博士后研究。2024年入选湖北省高层次人才计划，主要从事随机动力系统, 机器学习与动力系统相关研究。已在SIAM Journal on Scientific Computing, Nature Computational Science, CMAME, Physica D等期刊发表多篇学术论文。

邀请人：王季红