NEO: Non Equilibrium Sampling on the Orbit of a Deterministic Transform - Archive ouverte HAL
Conference Papers Year : 2021

NEO: Non Equilibrium Sampling on the Orbit of a Deterministic Transform

Abstract

Sampling from a complex distribution $\pi$ and approximating its intractable normalizing constant Z are challenging problems. In this paper, a novel family of importance samplers (IS) and Markov chain Monte Carlo (MCMC) samplers is derived. Given an invertible map T, these schemes combine (with weights) elements from the forward and backward Orbits through points sampled from a proposal distribution $\rho$. The map T does not leave the target $\pi$ invariant, hence the name NEO, standing for Non-Equilibrium Orbits. NEO-IS provides unbiased estimators of the normalizing constant and self-normalized IS estimators of expectations under $\pi$ while NEO-MCMC combines multiple NEO-IS estimates of the normalizing constant and an iterated sampling-importance resampling mechanism to sample from $\pi$. For T chosen as a discrete-time integrator of a conformal Hamiltonian system, NEO-IS achieves state-of-the art performance on difficult benchmarks and NEO-MCMC is able to explore highly multimodal targets. Additionally, we provide detailed theoretical results for both methods. In particular, we show that NEO-MCMC is uniformly geometrically ergodic and establish explicit mixing time estimates under mild conditions.
Fichier principal
Vignette du fichier
neo_arxiv.pdf (4.61 Mo) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

hal-03168489 , version 1 (13-03-2021)
hal-03168489 , version 2 (20-08-2021)

Identifiers

Cite

Achille Thin, Yazid Janati, Sylvain Le Corff, Charles Ollion, Arnaud Doucet, et al.. NEO: Non Equilibrium Sampling on the Orbit of a Deterministic Transform. 35th International Conference on Neural Information Processing Systems, Dec 2021, Online, France. ⟨10.5555/3540261.3541565⟩. ⟨hal-03168489v2⟩
646 View
250 Download

Altmetric

Share

More