GTH算法、删失马尔可夫链和$RG$ -因子分解
GTH Algorithm, Censored Markov Chains, and $RG$-Factorization
摘要:
GTH算法是计算马尔可夫链平稳概率的一种数值稳定算法。在数学上,GTH算法是高斯消去法的重排,因此它们在数学上是等价的。GTH算法中的所有组件都可以基于删失概念进行概率解释,并且GTH算法中的每个消除都会导致删失的马尔可夫链。$RG$ -因子分解是高斯消去法的LU-分解的对应物。对于无穷多个线性方程组成的系统,删失马尔可夫链也可以看作是GTH算法的一个扩展版本。在$l_1$ -范数下,删失马尔可夫链对原链的逼近误差最小.
英文摘要:
In this paper, we provide a review on the GTH algorithm, which is a numerically stable algorithm for computing stationary probabilities of a Markov chain. Mathematically the GTH algorithm is an rearrangement of Gaussian elimination, and therefore they are mathematically equivalent. All components in the GTH algorithm can be interpreted probabilistically based on the censoring concept and each elimination in the GTH algorithm leads to a censored Markov chain. The $RG$-factorization is a counterpart to the LU-decomposition for Gaussian elimination. The censored Markov chain can also be treated as an extended version of the GTH algorithm for a system consisting of infinitely many linear equations. The censored Markov chain produces a minimal error for approximating the original chain under the $l_1$-norm.
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。