Scalable Bayesian Inference For Generalized Multivariate Time-Series Models
My work as a statistical researcher with Prof. Justin Silverman at Penn State University towards submission to 39th AAAI Conference on Artificial Intelligence
Part 1: Parallelized Metropolis Hastings for Particle Refinement
Developed a high-performance parallelized Metropolis Particle Filter header library, meticulously engineered in C++. This library boasts cross-language interfaces to broaden accessibility, with a demonstrative interface implemented for R via the Rcpp package. Leveraging the Eigen library, it ensures efficient matrix algebra operations, while parallelization is achieved through OpenMP, optimizing performance across various computing environments. Engineered a high-performance header library for parallelized Metropolis Particle Filter in C++ for GDLM's posterior computation and integrated it with R via the Rcpp package for enhanced accessibility. This achieves a 32x faster performance than basic R implementation.