BriefGPT.xyz
Nov, 2017
利用神经网络泛化哈密顿蒙特卡罗
Generalizing Hamiltonian Monte Carlo with Neural Networks
HTML
PDF
Daniel Levy, Matthew D. Hoffman, Jascha Sohl-Dickstein
TL;DR
本文介绍了一种使用深度神经网络参数化的通用方法来训练Markov链蒙特卡洛核,该方法收敛快、混合快,并且我们在一系列简单但具有挑战性的分布中展示了大量的实证收益,并在一个真实的任务中展示了定量和定性的增益:潜变量生成建模。同时,我们还发布了算法的开源TensorFlow实现。
Abstract
We present a general-purpose method to train
markov chain monte carlo
kernels, parameterized by
deep neural networks
, that converge and mix quickly to their target distribution. Our method generalizes
→