Search papers, labs, and topics across Lattice.
This paper provides a theoretical comparison of two No-U-Turn Sampler (NUTS) variants, NUTS-mul and NUTS-BPS, by deriving necessary and sufficient conditions for geometric ergodicity. The authors establish the first sufficient conditions for geometric ergodicity and ergodicity for NUTS-mul and derive the first mixing time result for NUTS-BPS under a Gaussian target. The analysis reveals that while both variants exhibit similar qualitative behavior regarding geometric ergodicity, NUTS-BPS demonstrates faster convergence rates, with mixing times scaling as O(d^{1/4}) with smaller constants compared to NUTS-mul when initialized in the typical set of a Gaussian measure.
Despite their similar theoretical guarantees, NUTS-BPS converges faster than NUTS-mul, offering a potential efficiency boost for high-dimensional Bayesian inference.
The No-U-Turn Sampler (NUTS) is the computational workhorse of modern Bayesian software libraries, yet its qualitative and quantitative convergence guarantees were established only recently. A significant gap remains in the theoretical comparison of its two main variants: NUTS-mul and NUTS-BPS, which use multinomial sampling and biased progressive sampling, respectively, for index selection. In this paper, we address this gap in three contributions. First, we derive the first necessary conditions for geometric ergodicity for both variants. Second, we establish the first sufficient conditions for geometric ergodicity and ergodicity for NUTS-mul. Third, we obtain the first mixing time result for NUTS-BPS on a standard Gaussian distribution. Our results show that NUTS-mul and NUTS-BPS exhibit nearly identical qualitative behavior, with geometric ergodicity depending on the tail properties of the target distribution. However, they differ quantitatively in their convergence rates. More precisely, when initialized in the typical set of the canonical Gaussian measure, the mixing times of both NUTS-mul and NUTS-BPS scale as $O(d^{1/4})$ up to logarithmic factors, where $d$ denotes the dimension. Nevertheless, the associated constants are strictly smaller for NUTS-BPS.
