BriefGPT.xyz
Oct, 2018
FFJORD: 可伸缩的可逆生成模型的自由连续动力学
FFJORD: Free-form Continuous Dynamics for Scalable Reversible Generative Models
HTML
PDF
Will Grathwohl, Ricky T. Q. Chen, Jesse Betterncourt, Ilya Sutskever, David Duvenaud
TL;DR
本研究利用可逆神经网络将从简单分布映射到复杂分布的一类生成模型进行了优化，实现了无偏密度估计和单次采样，并在高维度密度估计、图像生成和变分推断等领域超越了目前的所有精确似然方法。
Abstract
A promising class of
generative models
maps points from a simple distribution to a complex distribution through an invertible
neural network
. Likelihood-based training of these models requires restricting their a
→