{"title":"用于动态系统参数估计的模型嵌入式高斯过程回归","authors":"Ying Zhou, Jinglai Li, Xiang Zhou, Hongqiao Wang","doi":"arxiv-2409.11745","DOIUrl":null,"url":null,"abstract":"Identifying dynamical system (DS) is a vital task in science and engineering.\nTraditional methods require numerous calls to the DS solver, rendering\nlikelihood-based or least-squares inference frameworks impractical. For\nefficient parameter inference, two state-of-the-art techniques are the kernel\nmethod for modeling and the \"one-step framework\" for jointly inferring unknown\nparameters and hyperparameters. The kernel method is a quick and\nstraightforward technique, but it cannot estimate solutions and their\nderivatives, which must strictly adhere to physical laws. We propose a\nmodel-embedded \"one-step\" Bayesian framework for joint inference of unknown\nparameters and hyperparameters by maximizing the marginal likelihood. This\napproach models the solution and its derivatives using Gaussian process\nregression (GPR), taking into account smoothness and continuity properties, and\ntreats differential equations as constraints that can be naturally integrated\ninto the Bayesian framework in the linear case. Additionally, we prove the\nconvergence of the model-embedded Gaussian process regression (ME-GPR) for\ntheoretical development. Motivated by Taylor expansion, we introduce a\npiecewise first-order linearization strategy to handle nonlinear dynamic\nsystems. We derive estimates and confidence intervals, demonstrating that they\nexhibit low bias and good coverage properties for both simulated models and\nreal data.","PeriodicalId":501215,"journal":{"name":"arXiv - STAT - Computation","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2024-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Model-Embedded Gaussian Process Regression for Parameter Estimation in Dynamical System\",\"authors\":\"Ying Zhou, Jinglai Li, Xiang Zhou, Hongqiao Wang\",\"doi\":\"arxiv-2409.11745\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Identifying dynamical system (DS) is a vital task in science and engineering.\\nTraditional methods require numerous calls to the DS solver, rendering\\nlikelihood-based or least-squares inference frameworks impractical. For\\nefficient parameter inference, two state-of-the-art techniques are the kernel\\nmethod for modeling and the \\\"one-step framework\\\" for jointly inferring unknown\\nparameters and hyperparameters. The kernel method is a quick and\\nstraightforward technique, but it cannot estimate solutions and their\\nderivatives, which must strictly adhere to physical laws. We propose a\\nmodel-embedded \\\"one-step\\\" Bayesian framework for joint inference of unknown\\nparameters and hyperparameters by maximizing the marginal likelihood. This\\napproach models the solution and its derivatives using Gaussian process\\nregression (GPR), taking into account smoothness and continuity properties, and\\ntreats differential equations as constraints that can be naturally integrated\\ninto the Bayesian framework in the linear case. Additionally, we prove the\\nconvergence of the model-embedded Gaussian process regression (ME-GPR) for\\ntheoretical development. Motivated by Taylor expansion, we introduce a\\npiecewise first-order linearization strategy to handle nonlinear dynamic\\nsystems. We derive estimates and confidence intervals, demonstrating that they\\nexhibit low bias and good coverage properties for both simulated models and\\nreal data.\",\"PeriodicalId\":501215,\"journal\":{\"name\":\"arXiv - STAT - Computation\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-09-18\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"arXiv - STAT - Computation\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/arxiv-2409.11745\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"arXiv - STAT - Computation","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2409.11745","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Model-Embedded Gaussian Process Regression for Parameter Estimation in Dynamical System
Identifying dynamical system (DS) is a vital task in science and engineering.
Traditional methods require numerous calls to the DS solver, rendering
likelihood-based or least-squares inference frameworks impractical. For
efficient parameter inference, two state-of-the-art techniques are the kernel
method for modeling and the "one-step framework" for jointly inferring unknown
parameters and hyperparameters. The kernel method is a quick and
straightforward technique, but it cannot estimate solutions and their
derivatives, which must strictly adhere to physical laws. We propose a
model-embedded "one-step" Bayesian framework for joint inference of unknown
parameters and hyperparameters by maximizing the marginal likelihood. This
approach models the solution and its derivatives using Gaussian process
regression (GPR), taking into account smoothness and continuity properties, and
treats differential equations as constraints that can be naturally integrated
into the Bayesian framework in the linear case. Additionally, we prove the
convergence of the model-embedded Gaussian process regression (ME-GPR) for
theoretical development. Motivated by Taylor expansion, we introduce a
piecewise first-order linearization strategy to handle nonlinear dynamic
systems. We derive estimates and confidence intervals, demonstrating that they
exhibit low bias and good coverage properties for both simulated models and
real data.