Saifuddin Syed, Alexandre Bouchard-Côté, Kevin Chern, Arnaud Doucet
{"title":"优化的校正序列蒙特卡罗采样器","authors":"Saifuddin Syed, Alexandre Bouchard-Côté, Kevin Chern, Arnaud Doucet","doi":"arxiv-2408.12057","DOIUrl":null,"url":null,"abstract":"Annealed Sequential Monte Carlo (SMC) samplers are special cases of SMC\nsamplers where the sequence of distributions can be embedded in a smooth path\nof distributions. Using this underlying path of distributions and a performance\nmodel based on the variance of the normalisation constant estimator, we\nsystematically study dense schedule and large particle limits. From our theory\nand adaptive methods emerges a notion of global barrier capturing the inherent\ncomplexity of normalisation constant approximation under our performance model.\nWe then turn the resulting approximations into surrogate objective functions of\nalgorithm performance, and use them for methodology development. We obtain\nnovel adaptive methodologies, Sequential SMC (SSMC) and Sequential AIS (SAIS)\nsamplers, which address practical difficulties inherent in previous adaptive\nSMC methods. First, our SSMC algorithms are predictable: they produce a\nsequence of increasingly precise estimates at deterministic and known times.\nSecond, SAIS, a special case of SSMC, enables schedule adaptation at a memory\ncost constant in the number of particles and require much less communication.\nFinally, these characteristics make SAIS highly efficient on GPUs. We develop\nan open-source, high-performance GPU implementation based on our methodology\nand demonstrate up to a hundred-fold speed improvement compared to\nstate-of-the-art adaptive AIS methods.","PeriodicalId":501215,"journal":{"name":"arXiv - STAT - Computation","volume":"230 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2024-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Optimised Annealed Sequential Monte Carlo Samplers\",\"authors\":\"Saifuddin Syed, Alexandre Bouchard-Côté, Kevin Chern, Arnaud Doucet\",\"doi\":\"arxiv-2408.12057\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Annealed Sequential Monte Carlo (SMC) samplers are special cases of SMC\\nsamplers where the sequence of distributions can be embedded in a smooth path\\nof distributions. Using this underlying path of distributions and a performance\\nmodel based on the variance of the normalisation constant estimator, we\\nsystematically study dense schedule and large particle limits. From our theory\\nand adaptive methods emerges a notion of global barrier capturing the inherent\\ncomplexity of normalisation constant approximation under our performance model.\\nWe then turn the resulting approximations into surrogate objective functions of\\nalgorithm performance, and use them for methodology development. We obtain\\nnovel adaptive methodologies, Sequential SMC (SSMC) and Sequential AIS (SAIS)\\nsamplers, which address practical difficulties inherent in previous adaptive\\nSMC methods. First, our SSMC algorithms are predictable: they produce a\\nsequence of increasingly precise estimates at deterministic and known times.\\nSecond, SAIS, a special case of SSMC, enables schedule adaptation at a memory\\ncost constant in the number of particles and require much less communication.\\nFinally, these characteristics make SAIS highly efficient on GPUs. We develop\\nan open-source, high-performance GPU implementation based on our methodology\\nand demonstrate up to a hundred-fold speed improvement compared to\\nstate-of-the-art adaptive AIS methods.\",\"PeriodicalId\":501215,\"journal\":{\"name\":\"arXiv - STAT - Computation\",\"volume\":\"230 1\",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-08-22\",\"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-2408.12057\",\"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-2408.12057","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Optimised Annealed Sequential Monte Carlo Samplers
Annealed Sequential Monte Carlo (SMC) samplers are special cases of SMC
samplers where the sequence of distributions can be embedded in a smooth path
of distributions. Using this underlying path of distributions and a performance
model based on the variance of the normalisation constant estimator, we
systematically study dense schedule and large particle limits. From our theory
and adaptive methods emerges a notion of global barrier capturing the inherent
complexity of normalisation constant approximation under our performance model.
We then turn the resulting approximations into surrogate objective functions of
algorithm performance, and use them for methodology development. We obtain
novel adaptive methodologies, Sequential SMC (SSMC) and Sequential AIS (SAIS)
samplers, which address practical difficulties inherent in previous adaptive
SMC methods. First, our SSMC algorithms are predictable: they produce a
sequence of increasingly precise estimates at deterministic and known times.
Second, SAIS, a special case of SSMC, enables schedule adaptation at a memory
cost constant in the number of particles and require much less communication.
Finally, these characteristics make SAIS highly efficient on GPUs. We develop
an open-source, high-performance GPU implementation based on our methodology
and demonstrate up to a hundred-fold speed improvement compared to
state-of-the-art adaptive AIS methods.