The article presents the SGRLD method as a novel approach for efficiently estimating spatial covariance parameters in large datasets. By employing stochastic gradient Langevin dynamics (SGRLD), the method uses advanced techniques such as momentum and gradient information to aid in convergence and exploration of the posterior distribution. A comparative simulation study tested its performance against state-of-the-art Bayesian methods, utilizing metrics like Mean Squared Error and effective sample sizes to demonstrate SGRLD's advantages in computational efficiency and estimation accuracy.
In our study, we introduced the SGRLD method to efficiently estimate spatial covariance parameters and benchmarked it against state-of-the-art Bayesian methods.
Our evaluation metrics for the SGRLD included Mean Squared Error and effective sample sizes, demonstrating its superior performance in computational efficiency.
Collection
[
|
...
]