AN ACCELERATION STRATEGY FOR RANDOMIZE-THEN-OPTIMIZE SAMPLING VIA DEEP NEURAL NETWORKS*

doi:10.4208/jcm.2102-m2020-0339

CSpace

	AN ACCELERATION STRATEGY FOR RANDOMIZE-THEN-OPTIMIZE SAMPLING VIA DEEP NEURAL NETWORKS*
	Yan, Liang 1,2; Zhou, Tao 3
	2021
发表期刊	JOURNAL OF COMPUTATIONAL MATHEMATICS
ISSN	0254-9409
卷号	39 期号:6 页码:848-864
摘要	Randomize-then-optimize (RTO) is widely used for sampling from posterior distribu-tions in Bayesian inverse problems. However, RTO can be computationally intensive for complexity problems due to repetitive evaluations of the expensive forward model and its gradient. In this work, we present a novel goal-oriented deep neural networks (DNN) sur-rogate approach to substantially reduce the computation burden of RTO. In particular, we propose to drawn the training points for the DNN-surrogate from a local approximated posterior distribution - yielding a flexible and efficient sampling algorithm that converges to the direct RTO approach. We present a Bayesian inverse problem governed by elliptic PDEs to demonstrate the computational accuracy and efficiency of our DNN-RTO ap-proach, which shows that DNN-RTO can significantly outperform the traditional RTO.
关键词	Bayesian inverse problems Deep neural network Markov chain Monte Carlo
DOI	10.4208/jcm.2102-m2020-0339
收录类别	SCI
语种	英语
资助项目	NSF of China[11771081] ; NSF of China[11822111] ; NSF of China[11688101] ; NSF of China[11731006] ; science challenge project, China[TZ2018001] ; Zhishan Young Scholar Program of SEU, China ; National Key R&D Program of China[2020YFA0712000] ; Strategic Priority Research Program of Chinese Academy of Sciences[XDA25000404] ; youth innovation promotion association (CAS), China ; science challenge project[TZ2018001]
WOS研究方向	Mathematics
WOS类目	Mathematics, Applied ; Mathematics
WOS记录号	WOS:000711024000003
出版者	GLOBAL SCIENCE PRESS
引用统计
文献类型	期刊论文
条目标识符	http://ir.amss.ac.cn/handle/2S8OKBNM/59478
专题	中国科学院数学与系统科学研究院
通讯作者	Zhou, Tao
作者单位	1.Southeast Univ, Sch Math, Nanjing 210096, Peoples R China 2.Nanjing Ctr Appl Math, Nanjing 211135, Peoples R China 3.Chinese Acad Sci, Acad Math & Syst Sci, LSEC, Inst Computat Math & Sci Engn Comp, Beijing 100190, Peoples R China
推荐引用方式 GB/T 7714	Yan, Liang,Zhou, Tao. AN ACCELERATION STRATEGY FOR RANDOMIZE-THEN-OPTIMIZE SAMPLING VIA DEEP NEURAL NETWORKS*[J]. JOURNAL OF COMPUTATIONAL MATHEMATICS,2021,39(6):848-864.
APA	Yan, Liang,&Zhou, Tao.(2021).AN ACCELERATION STRATEGY FOR RANDOMIZE-THEN-OPTIMIZE SAMPLING VIA DEEP NEURAL NETWORKS*.JOURNAL OF COMPUTATIONAL MATHEMATICS,39(6),848-864.
MLA	Yan, Liang,et al."AN ACCELERATION STRATEGY FOR RANDOMIZE-THEN-OPTIMIZE SAMPLING VIA DEEP NEURAL NETWORKS*".JOURNAL OF COMPUTATIONAL MATHEMATICS 39.6(2021):848-864.