报告题目:AI for Stochastic Physics
报告时间:2024年11月22日 15:00
报告地点:成龙校区第一教学楼A107
报告摘要:
Machine learning and stochastic dynamics have deep connections and cross-feed each other. I will report our recent progress in three specific questions: (1) tracking the time-evolving probability distribution for stochastic reaction networks by the variational autoregressive neural network; (2) characterizing a type of dynamical phase transition in nonequilibrium statistical mechanics; (3) learning noise-induced transitions by multi-scaling reservoir computing. The results demonstrate how machine learning can help understand stochastic dynamics.
报告人简介:
汤迎,电子科技大学基础与前沿研究院教授,国家级青年人才,研究领域为机器学习、统计物理、开放量子系统、随机动力学等。近期研究成果包括:提出了演化神经网络的方法追踪高维随机反应网络演化;发展了时间序列动力学互信息的计算框架;发现非平衡量子系统中磁场不做功却仍能增大自由能等。研究发表在Nature Machine Intelligence,Nature Communications,Physics Review E,Nature,Nature Methods等,共发表论文23篇,其中第一或通讯作者16篇,谷歌学术引用3800余次。
【编辑:唐荣】
(微信扫描分享)