强烈推荐soVeryTired · 2017-02
“Radford Neal's expository paper [1] on HMC is a beautiful overview of the area, and honestly one of the most enlightening papers I've ever read.”
强烈推荐 Neal 这篇 HMC 综述,称是读过最有启发的论文之一
Neal 对哈密顿蒙特卡洛(HMC)原理、实践与变体的经典综述
哈密顿动力学可为 Metropolis 算法产生远距离提议,避免简单随机游走提议那种扩散式的缓慢探索;通过引入虚构的「动量」变量,它可用于大多数连续状态空间问题。其关键在于哈密顿动力学保持体积,轨迹因此可用于定义复杂映射而无需计算难算的雅可比因子,即使时间被离散化近似也能精确保持。作者在综述中讨论了 HMC 的理论与实践,并介绍多种变体,如用状态窗口决定接受或拒绝、用快速近似计算轨迹、在轨迹中做回火以处理孤立模式,以及避免无用轨迹耗费算力的捷径方法。
评论者称它是 HMC 的优美概览,甚至是读过的最有启发的论文之一,并向想学习采样、MCMC 的人推荐。
“Radford Neal's expository paper [1] on HMC is a beautiful overview of the area, and honestly one of the most enlightening papers I've ever read.”
“Markov Chain Monte Carlo to sample from probability distributions is a good start - MCMC using Hamiltonian dynamics if you are into sampling.”
“Another commenter mentioned HMC [3], which is just a specific instance of MCMC.”