GPU Accelerated Nested Sampling

Nested Sampling is a powerful Monte Carlo method for Bayesian inference, but is often viewed as a legacy, CPU-bound algorithm. We reformulate Nested Sampling for modern GPU hardware, leveraging vectorization and a slice sampling-based inner kernel. Our implementation in the blackjax framework achieves state-of-the-art performance, enabling rapid and robust parameter estimation and model comparison for challenging problems in gravitational wave astronomy and cosmology.