1

Hamiltonian Monte Carlo算法

 2 years ago
source link: https://blog.xpgreat.com/p/hmc/
Go to the source link to view the article. You can view the picture content, updated content and better typesetting reading experience. If the link is broken, please click the button below to view the snapshot at that time.
neoserver,ios ssh client

MCMC(Markov-Chain Monte-Carlo)算法用于生成给定分布的样本。在很多时候,我们想计算一个复杂分布的函数期望值,但用解析的,求解积分的方法,是极其困难的,对于有些分布甚至是不可能的。所以我们使用一些采样方法,例如Gibbs sampling,Metropolis-Hasting算法。本文介绍Hamiltonian Monte-Carlo算法。

目标和问题

给定一个分布π(x)\pi(x)π(x),计算期望:Eπ[f]=∫f(q)π(q)dq\mathbb E_\pi[f]=\int f(q)\pi(q)dqEπ​[f]=∫f(q)π(q)dq,而直接计算积分是不可行的。因此我们计算期望的一个近似:Eπ[f]≈1N∑if(xi)\mathbb E_\pi[f] \approx \frac{1}{N}\sum_i f(x_i)Eπ​[f]≈N1​∑i​f(xi​)

Licensed under CC BY-NC-SA 4.0


About Joyk


Aggregate valuable and interesting links.
Joyk means Joy of geeK