2019/11/22 by Paul Brown, Chaitanya Joshi, Brown, Paul T. +4 · 1 citation
Decision Sciences · Environmental Science · Physics and Astronomy · #Computation (stat.CO) #FOS: Computer and information sciences #Model Reduction and Neural Networks #Probabilistic and Robust Engineering Design #Soil Geostatistics and Mapping
paper · pdf · doi:10.48550/arxiv.1911.09880
openalex publication_date 2019/11/22 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28
Recently, it has been shown that approximations to marginal posterior\ndistributions obtained using a low discrepancy sequence (LDS) can outperform\nstandard grid-based methods with respect to both accuracy and computational\nefficiency. This recent method, which we will refer to as LDS-StM, can also\nproduce good approximations to multimodal posteriors. However, implementation\nof LDS-StM into integrated nested Laplace approximations (INLA), a methodology\nin which grid-based methods are used, is challenging. Motivated by this\nproblem, we propose modifications to LDS-StM that improves the approximations\nand make it compatible with INLA, without sacrificing computational speed. We\nalso present two examples to demonstrate that LDS-StM with modifications can\noutperform INLA's own grid approximation with respect to speed and accuracy. We\nalso demonstrate the flexibility of the new approach for the approximation of\nmultimodal marginals.\n