基于机器学习由暂态数据预测系统的演化
摘要:
如何通过少量的暂态数据去预测系统的长时间的动力学行为,是一个重要问题。对Sturat-Landau系 统、布鲁塞尔振子以及 Belousov-Zhabotinsky系统的稳态数据进行采集,并输入到储备池神经网络里进行训练;然后基于3个系统采集的少量的暂态数据,利用已经训练好的神经网络模型,能够对其在不同参数下长时间的动力学行 为精准地预测。研究结果有利于加深对复杂系统如何对外来的变化或扰动做出反应的理解 。
Predieting the dynamie behavior of a system based on transient data from unknown dynamic equations is an important problem. Firstly, we colleet steady-state data from the $tuart-landau system, the Brusselator, and the Belousov-Zhabotinsky system, and put them into a reservoir neural network for training. Then, basing on the transient data collected from the three systems, and using the trained neural network model, we can accurately prediet their long-term dynamic behav-ior under different parameters, The findings of this study play an important role in understanding how systems respond to ex-ternal changes, disturbances, or inputs, thus highlighting the significance of studying transient dynamics.
作者:
王江峰,陈银霞,祁月盈,姚成贵
Wang Jiangfeng , Chen Yinxia, Qi Yueying,Yao Chenggui
机构地区:
浙江师范大学数学学院;嘉兴大学教务处;数据科学学院
引用本文:
王江峰,陈银霞,祁月盈等。基于机器学习由暂态数据预测系统的演化[J].betway官方app 学报(自然科学版),2024,52(6):113-118.
(WangJiangfeng,Chen Yinxia,QiYueying,etal.Predicting the evolution ofsystemsbased on a transientdata with machine learning[J] . Journal of Henan Normal University(Natural Science Edition) , 2024,52(6) :113-118. DOI:10. 16366/j. cnki.1000-2367. 2024. 03. 20. 0001. )
基金:
国家自然科学基金;浙江省自然科学基金
关键词:
储备池神经网络;动力学行为;Stuart-Landau系统;布鲁塞尔振子;BZ反应系统
reservoir computing neural net works; dynamie behavior; Stuart-landau system; Brusselator; BelousovZhabotinskyreactionsystem
分类号:
O415. 6;N93;O244