Decomposed slice sampling for factorized distributions

作者：

Highlights：

• We present an MCMC sampling framework called decomposed slice sampling.

• DSS treats the factorized target distribution as two multipliers.

• The two multipliers are used in two successive steps to implement effective sampling.

• The proposed model outperforms multiple baselines for sampling complex distributions.

摘要

•We present an MCMC sampling framework called decomposed slice sampling.•DSS treats the factorized target distribution as two multipliers.•The two multipliers are used in two successive steps to implement effective sampling.•The proposed model outperforms multiple baselines for sampling complex distributions.

论文关键词：Slice sampling,Markov chain Monte Carlo,Decomposed slice sampling,Hamiltonian Monte Carlo

论文评审过程：Received 27 September 2018, Revised 19 July 2019, Accepted 25 August 2019, Available online 26 August 2019, Version of Record 30 August 2019.

论文官网地址：https://doi.org/10.1016/j.patcog.2019.107021