Stratified Multiple-descendant Growth (SMG)

Introduction

We demonstrate the SMG method proposed in Section 5 of Li, Wang, Deng and Liu (2020). SMG allows the sequential Monte Carlo (SMC) to explore the sample space more efficiently compared to the standard i.i.d. multiple-descendant sampling-resampling approach.

We provide three notebook tutorials on terrain navigation, estimation of multivariate stochastic volatility and sampling from a high-dimensional target distribution. The details of the simulation setttings can be found in Section 6 and Appendix B of the paper.

Reference

Yichao Li, Wenshuo Wang, Ke Deng and Jun Liu. (2020). Stratification and Optimal Resampling for Sequential Monte Carlo. [pdf] [arXiv] [code]