Minglei Yang, Pengjun Wang, Ming Fan, Dan Lu, Yanzhao Cao, Guannan Zhang
{"title":"量化不确定性传播中代用模型的条件伪可逆归一化流程","authors":"Minglei Yang, Pengjun Wang, Ming Fan, Dan Lu, Yanzhao Cao, Guannan Zhang","doi":"arxiv-2404.00502","DOIUrl":null,"url":null,"abstract":"We introduce a conditional pseudo-reversible normalizing flow for\nconstructing surrogate models of a physical model polluted by additive noise to\nefficiently quantify forward and inverse uncertainty propagation. Existing\nsurrogate modeling approaches usually focus on approximating the deterministic\ncomponent of physical model. However, this strategy necessitates knowledge of\nnoise and resorts to auxiliary sampling methods for quantifying inverse\nuncertainty propagation. In this work, we develop the conditional\npseudo-reversible normalizing flow model to directly learn and efficiently\ngenerate samples from the conditional probability density functions. The\ntraining process utilizes dataset consisting of input-output pairs without\nrequiring prior knowledge about the noise and the function. Our model, once\ntrained, can generate samples from any conditional probability density\nfunctions whose high probability regions are covered by the training set.\nMoreover, the pseudo-reversibility feature allows for the use of\nfully-connected neural network architectures, which simplifies the\nimplementation and enables theoretical analysis. We provide a rigorous\nconvergence analysis of the conditional pseudo-reversible normalizing flow\nmodel, showing its ability to converge to the target conditional probability\ndensity function using the Kullback-Leibler divergence. To demonstrate the\neffectiveness of our method, we apply it to several benchmark tests and a\nreal-world geologic carbon storage problem.","PeriodicalId":501061,"journal":{"name":"arXiv - CS - Numerical Analysis","volume":"116 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2024-03-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Conditional Pseudo-Reversible Normalizing Flow for Surrogate Modeling in Quantifying Uncertainty Propagation\",\"authors\":\"Minglei Yang, Pengjun Wang, Ming Fan, Dan Lu, Yanzhao Cao, Guannan Zhang\",\"doi\":\"arxiv-2404.00502\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"We introduce a conditional pseudo-reversible normalizing flow for\\nconstructing surrogate models of a physical model polluted by additive noise to\\nefficiently quantify forward and inverse uncertainty propagation. Existing\\nsurrogate modeling approaches usually focus on approximating the deterministic\\ncomponent of physical model. However, this strategy necessitates knowledge of\\nnoise and resorts to auxiliary sampling methods for quantifying inverse\\nuncertainty propagation. In this work, we develop the conditional\\npseudo-reversible normalizing flow model to directly learn and efficiently\\ngenerate samples from the conditional probability density functions. The\\ntraining process utilizes dataset consisting of input-output pairs without\\nrequiring prior knowledge about the noise and the function. Our model, once\\ntrained, can generate samples from any conditional probability density\\nfunctions whose high probability regions are covered by the training set.\\nMoreover, the pseudo-reversibility feature allows for the use of\\nfully-connected neural network architectures, which simplifies the\\nimplementation and enables theoretical analysis. We provide a rigorous\\nconvergence analysis of the conditional pseudo-reversible normalizing flow\\nmodel, showing its ability to converge to the target conditional probability\\ndensity function using the Kullback-Leibler divergence. To demonstrate the\\neffectiveness of our method, we apply it to several benchmark tests and a\\nreal-world geologic carbon storage problem.\",\"PeriodicalId\":501061,\"journal\":{\"name\":\"arXiv - CS - Numerical Analysis\",\"volume\":\"116 1\",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-03-31\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"arXiv - CS - Numerical Analysis\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/arxiv-2404.00502\",\"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 - CS - Numerical Analysis","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2404.00502","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Conditional Pseudo-Reversible Normalizing Flow for Surrogate Modeling in Quantifying Uncertainty Propagation
We introduce a conditional pseudo-reversible normalizing flow for
constructing surrogate models of a physical model polluted by additive noise to
efficiently quantify forward and inverse uncertainty propagation. Existing
surrogate modeling approaches usually focus on approximating the deterministic
component of physical model. However, this strategy necessitates knowledge of
noise and resorts to auxiliary sampling methods for quantifying inverse
uncertainty propagation. In this work, we develop the conditional
pseudo-reversible normalizing flow model to directly learn and efficiently
generate samples from the conditional probability density functions. The
training process utilizes dataset consisting of input-output pairs without
requiring prior knowledge about the noise and the function. Our model, once
trained, can generate samples from any conditional probability density
functions whose high probability regions are covered by the training set.
Moreover, the pseudo-reversibility feature allows for the use of
fully-connected neural network architectures, which simplifies the
implementation and enables theoretical analysis. We provide a rigorous
convergence analysis of the conditional pseudo-reversible normalizing flow
model, showing its ability to converge to the target conditional probability
density function using the Kullback-Leibler divergence. To demonstrate the
effectiveness of our method, we apply it to several benchmark tests and a
real-world geologic carbon storage problem.