arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2301.07388stat.MLcs.LG
收藏

学习玻尔兹曼密度之间的插值

Learning Interpolations between Boltzmann Densities

University of Geneva(日内瓦大学)

Bálint Máté, François Fleuret

AI总结:

本文提出一种无需样本、仅依赖能量函数的连续归一化流训练目标,通过学习能量函数插值并优化向量场满足连续性方程,实现玻尔兹曼密度间的传输。

摘要:

我们为连续归一化流引入一种训练目标,该目标可在没有样本但存在能量函数的情况下使用。我们的方法依赖于目标能量 $f_1$ 与广义高斯能量函数 $f_0(x) = ||x/σ||_p^p$ 之间能量函数的指定或学习得到的插值 $f_t$。能量函数的插值诱导出玻尔兹曼密度的插值 $p_t \propto e^{-f_t}$,我们旨在找到一个时间相关向量场 $V_t$,使其沿密度族 $p_t$ 传输样本。沿密度族 $p_t$ 传输样本的条件等价于用 $V_t$ 和 $p_t = Z_t^{-1}e^{-f_t}$ 满足连续性方程。因此,我们优化 $V_t$ 和 $f_t$ 以满足该偏微分方程。我们在高斯混合模型和双阱势中量子力学粒子的玻尔兹曼密度上,对所提出的训练目标与反向 KL 散度进行了实验比较。

英文摘要:

We introduce a training objective for continuous normalizing flows that can be used in the absence of samples but in the presence of an energy function. Our method relies on either a prescribed or a learnt interpolation $f_t$ of energy functions between the target energy $f_1$ and the energy function of a generalized Gaussian $f_0(x) = ||x/σ||_p^p$. The interpolation of energy functions induces an interpolation of Boltzmann densities $p_t \propto e^{-f_t}$ and we aim to find a time-dependent vector field $V_t$ that transports samples along the family $p_t$ of densities. The condition of transporting samples along the family $p_t$ is equivalent to satisfying the continuity equation with $V_t$ and $p_t = Z_t^{-1}e^{-f_t}$. Consequently, we optimize $V_t$ and $f_t$ to satisfy this partial differential equation. We experimentally compare the proposed training objective to the reverse KL-divergence on Gaussian mixtures and on the Boltzmann density of a quantum mechanical particle in a double-well potential.

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。

↑