NP-NUTS: A Nonparametric No-U-Turn Sampler

Explore a 16-minute conference talk from the LAFI 2025 workshop that introduces NP-NUTS, a novel hybrid algorithm combining the No-U-Turn Sampler (NUTS) and Nonparametric Hamiltonian Monte Carlo (NP-HMC). Presented by researchers from the University of Oxford, NTU, and Utrecht University, this video explains how NP-NUTS automatically adjusts trajectory lengths to create a dynamically tuned version of NP-HMC. Learn about the key challenges addressed in developing this algorithm, including ensuring correct sampling from the constructed tree and modifying stopping conditions to preserve detailed balance. The presentation was delivered at the LAFI workshop on January 19, 2025, sponsored by ACM SIGPLAN.