Randomized trial demonstrates high acceptance rates in sampling directional data on toroidal and spherical distributions, indicating efficient random variate generation.
Key Points
This research aims to create an efficient framework for sampling from various directional distributions, specifically extending techniques used in one-dimensional circular contexts to higher dimensions.
Developed a modified acceptance-rejection algorithm for sampling directional data.
Applied the approach to both toroidal and spherical distributions, demonstrating its efficiency compared to previous methods.
Introduced a von Mises-like distribution on curved torus, considering its geometry in R3.
Achieved consistently higher acceptance rates and lower runtimes than established sampling methods.
Successfully extended the algorithm to two-dimensional toroidal and spherical settings.
Demonstrated the ability to capture mean direction concentrations while considering curvature effects.