Bayesian model comparison via path-sampling sequential Monte Carlo
Bayesian model comparison via path-sampling sequential Monte Carlo
复制标题
DOI：
10.1109/ssp.2012.6319672
复制
发表时间：
2012-10
期刊：
影响因子：
--
通讯作者：
Yan Zhou;A. M. Johansen;J. Aston
中科院分区：
文献类型：
--
作者：
Yan Zhou;A. M. Johansen;J. Aston
为了选择、平均和验证的目的进行模型比较是贯穿统计和信号处理的问题。在贝叶斯范式中，这些问题都需要计算特定类别中模型的后验概率。近年来取得了重大进展，但在实际执行现有计划方面存在许多困难。本文开发了序贯蒙特卡罗（SMC）抽样策略，以表征后验分布的一组模型，以及这些模型的参数，这在一定程度上解决了使用其他技术所遇到的困难。特别地，仅使用模型模拟内提供感兴趣的量的有效表征。概念验证模拟证明了算法的鲁棒性和潜在性能，特别是通过使用GPU或多核处理器的并行实现。
Model comparison for the purposes of selection, averaging and validation is a problem which is found throughout statistics and signal processing. Within the Bayesian paradigm, these problems all require the calculation of the posterior probabilities of models within a particular class. Substantial progress has been made in recent years, but there are numerous difficulties in the practical implementation of existing schemes. This paper develops sequential Monte Carlo (SMC) sampling strategies to characterize the posterior distribution of a collection of models, as well as the parameters of those models which go some way towards addressing the difficulties encountered using other techniques. In particular, efficient characterization of the quantities of interest is provided using only within model simulation. Proof-of-concept simulations demonstrate the robustness and potential performance of the algorithm, particular via parallel implementation using a GPU or multi-core processor.