摘要
Sloppy属性是很多模拟实际系统的数学模型所共有的一个属性。在过去20年间,sloppy属性及相关特征的研究得到了研究者们的广泛关注。本文从sloppy属性的定义、应用、判定依据,到该属性的起因及消除方法进行了探讨,最后对sloppy模型中的参数估计及相关的影响因素进行了综述。Sloppy属性的存在为模型参数的认定及未来的模型预测提出了很大的挑战,因此本综述重点关注了研究者们对于sloppy属性的起源及消除的多角度探索。模型中sloppy参数的认定问题引发了对于sloppy属性的研究,从算法角度如何解决这个问题也是本综述的核心部分。除算法本身之外,对参数估计效果有影响或受参数估计结果影响的因素也一并进行了分析总结。通过本综述,我们给出了sloppy属性研究发展的概括性总结并显示了对该问题研究发展的清晰图示。最后也给出了未来关于sloppy属性及相关问题研究的几个有意思的主题。
Sloppy property is common for many mathematical models simulating real systems.In the past twenty years,the study on sloppy and related properties has received extensive attention from researchers.This paper gives a comprehensive review of sloppiness from the definition,applications,identification standards,causing and curing of it,to the parameter estimation methods and the influence factors.The existence of sloppy property poses a great challenge for parameter identification and future model prediction,so this review focused on the research on the origin and elimi-nation of it in many ways.The problem of sloppy parameter identification triggered the research of sloppy properties,so it becomes the main part of this review that how to solve it by algorithms.In addition to algorithms,influencing fac-tors and affected factors about parameter estimation are discussed together.Through this review,we summarized the development of sloppy properties and gave a clear picture of how it has advanced.Lastly,some future interesting topics about sloppy and related problems are given.
作者
王红
康玲
WANG Hong;KANG Ling(School of Computer,Dalian Neusoft University of Information,Dalian Liaoning 116023,China;School of Software,Dalian Neusoft University of Information,Dalian Liaoning 116023,China)
出处
《计算机仿真》
北大核心
2023年第2期6-14,176,共10页
Computer Simulation
基金
辽宁省教育厅面上项目《大尺度非线性计算昂贵优化问题关键技术研究》(LJKMZ20222004)
国家重点研发计划项目《基于人工智能的社会治理安全风险预警与智控关键技术装备研发与应用示范》(2021YFC3320300)资助。
关键词
凌乱属性
参数估计
起因及消除
影响因素
优化算法
Sloppiness
Parameter estimation
Causing and curing
Influence factors
Optimization algorithm
