HOPS: high-performance library for (non-)uniform sampling of convex-constrained models

Area of Science:

• Computational Mathematics
• Statistical Computing
• Software Development

Background:

• Sampling from complex probability distributions is crucial in various scientific domains.
• Existing methods for convex-constrained models can be computationally intensive and difficult to integrate.

Purpose of the Study:

• To introduce the Highly Optimized Polytope Sampling (HOPS) C++ library.
• To provide efficient and scalable algorithms for sampling convex-constrained models with arbitrary target functions.
• To demonstrate the utility of HOPS in Bayesian inference and its interoperability with other software.

Main Methods:

• Implementation of advanced sampling algorithms within a C++ library.
• Focus on uniform and non-uniform sampling techniques for polytope spaces.
• Integration with third-party software for practical applications, such as Bayesian inference.

Main Results:

• HOPS achieves substantial performance gains for uniform sampling compared to state-of-the-art methods.
• The library facilitates easy integration of non-uniform sampling.
• Demonstrated successful application in a Bayesian inference setting.

Conclusions:

• HOPS provides a powerful, efficient, and user-friendly tool for sampling convex-constrained models.
• The library enhances computational efficiency and simplifies complex sampling tasks.
• HOPS is a valuable asset for researchers in fields requiring advanced statistical sampling methods.