Verlet流:用于基于流的生成模型的精确似然积分器
Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models
AI总结:
本文提出Verlet流,利用辛积分器启发的Taylor-Verlet积分器在增广状态空间上构建精确似然的连续归一化流,解决了重要性采样中似然近似问题,实验表明其性能与autograd相当且速度更快。
摘要:
在使用连续归一化流(CNFs)计算模型似然时的近似问题,阻碍了这些模型在需要精确似然的玻尔兹曼分布重要性采样中的应用。在这项工作中,我们提出了Verlet流,这是一类在增广状态空间上的CNFs,其灵感来源于哈密顿动力学中的辛积分器。当与精心构造的Taylor-Verlet积分器一起使用时,Verlet流提供了精确似然的生成模型,这些模型在施加最小表达力约束的同时,将耦合流架构从非连续设置推广到连续设置。在玩具密度上的实验中,我们证明了常用的Hutchinson迹估计器的方差不适合重要性采样,而Verlet流在性能上与完整的autograd迹计算相当,同时速度显著更快。
英文摘要:
Approximations in computing model likelihoods with continuous normalizing flows (CNFs) hinder the use of these models for importance sampling of Boltzmann distributions, where exact likelihoods are required. In this work, we present Verlet flows, a class of CNFs on an augmented state-space inspired by symplectic integrators from Hamiltonian dynamics. When used with carefully constructed Taylor-Verlet integrators, Verlet flows provide exact-likelihood generative models which generalize coupled flow architectures from a non-continuous setting while imposing minimal expressivity constraints. On experiments over toy densities, we demonstrate that the variance of the commonly used Hutchinson trace estimator is unsuitable for importance sampling, whereas Verlet flows perform comparably to full autograd trace computations while being significantly faster.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。