{"title":"贝叶斯忠实数据同化的深度贝叶斯过滤器","authors":"Yuta Tarumi, Keisuke Fukuda, Shin-ichi Maeda","doi":"arxiv-2405.18674","DOIUrl":null,"url":null,"abstract":"State estimation for nonlinear state space models is a challenging task.\nExisting assimilation methodologies predominantly assume Gaussian posteriors on\nphysical space, where true posteriors become inevitably non-Gaussian. We\npropose Deep Bayesian Filtering (DBF) for data assimilation on nonlinear state\nspace models (SSMs). DBF constructs new latent variables $h_t$ on a new latent\n(``fancy'') space and assimilates observations $o_t$. By (i) constraining the\nstate transition on fancy space to be linear and (ii) learning a Gaussian\ninverse observation operator $q(h_t|o_t)$, posteriors always remain Gaussian\nfor DBF. Quite distinctively, the structured design of posteriors provides an\nanalytic formula for the recursive computation of posteriors without\naccumulating Monte-Carlo sampling errors over time steps. DBF seeks the\nGaussian inverse observation operators $q(h_t|o_t)$ and other latent SSM\nparameters (e.g., dynamics matrix) by maximizing the evidence lower bound.\nExperiments show that DBF outperforms model-based approaches and latent\nassimilation methods in various tasks and conditions.","PeriodicalId":501065,"journal":{"name":"arXiv - PHYS - Data Analysis, Statistics and Probability","volume":"56 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2024-05-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Deep Bayesian Filter for Bayes-faithful Data Assimilation\",\"authors\":\"Yuta Tarumi, Keisuke Fukuda, Shin-ichi Maeda\",\"doi\":\"arxiv-2405.18674\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"State estimation for nonlinear state space models is a challenging task.\\nExisting assimilation methodologies predominantly assume Gaussian posteriors on\\nphysical space, where true posteriors become inevitably non-Gaussian. We\\npropose Deep Bayesian Filtering (DBF) for data assimilation on nonlinear state\\nspace models (SSMs). DBF constructs new latent variables $h_t$ on a new latent\\n(``fancy'') space and assimilates observations $o_t$. By (i) constraining the\\nstate transition on fancy space to be linear and (ii) learning a Gaussian\\ninverse observation operator $q(h_t|o_t)$, posteriors always remain Gaussian\\nfor DBF. Quite distinctively, the structured design of posteriors provides an\\nanalytic formula for the recursive computation of posteriors without\\naccumulating Monte-Carlo sampling errors over time steps. DBF seeks the\\nGaussian inverse observation operators $q(h_t|o_t)$ and other latent SSM\\nparameters (e.g., dynamics matrix) by maximizing the evidence lower bound.\\nExperiments show that DBF outperforms model-based approaches and latent\\nassimilation methods in various tasks and conditions.\",\"PeriodicalId\":501065,\"journal\":{\"name\":\"arXiv - PHYS - Data Analysis, Statistics and Probability\",\"volume\":\"56 1\",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-05-29\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"arXiv - PHYS - Data Analysis, Statistics and Probability\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/arxiv-2405.18674\",\"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 - PHYS - Data Analysis, Statistics and Probability","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2405.18674","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Deep Bayesian Filter for Bayes-faithful Data Assimilation
State estimation for nonlinear state space models is a challenging task.
Existing assimilation methodologies predominantly assume Gaussian posteriors on
physical space, where true posteriors become inevitably non-Gaussian. We
propose Deep Bayesian Filtering (DBF) for data assimilation on nonlinear state
space models (SSMs). DBF constructs new latent variables $h_t$ on a new latent
(``fancy'') space and assimilates observations $o_t$. By (i) constraining the
state transition on fancy space to be linear and (ii) learning a Gaussian
inverse observation operator $q(h_t|o_t)$, posteriors always remain Gaussian
for DBF. Quite distinctively, the structured design of posteriors provides an
analytic formula for the recursive computation of posteriors without
accumulating Monte-Carlo sampling errors over time steps. DBF seeks the
Gaussian inverse observation operators $q(h_t|o_t)$ and other latent SSM
parameters (e.g., dynamics matrix) by maximizing the evidence lower bound.
Experiments show that DBF outperforms model-based approaches and latent
assimilation methods in various tasks and conditions.