Nagata Kenji, Watanabe Sumio
Department of Computational Intelligence and Systems Science, Tokyo Institute of Technology, Japan.
Neural Netw. 2008 Sep;21(7):980-8. doi: 10.1016/j.neunet.2007.11.002. Epub 2008 Feb 12.
The exchange Monte Carlo (EMC) algorithm is well known as being an improvement on the Markov Chain Monte Carlo method. Although it has been shown to be effective in many different contexts, the mathematical foundation of the EMC method has not yet been established. In this paper, we derive the asymptotic behavior of the symmetrized Kullback divergence and the exchange ratio, which is the acceptance ratio of the exchange process for the EMC method. In addition, based on these derived results, we propose optimal settings for the EMC method.
交换蒙特卡罗（EMC）算法是对马尔可夫链蒙特卡罗方法的一种改进，这一点广为人知。尽管它已在许多不同情况下被证明是有效的，但EMC方法的数学基础尚未确立。在本文中，我们推导了对称化库尔贝克散度和交换率的渐近行为，其中交换率是EMC方法中交换过程的接受率。此外，基于这些推导结果，我们提出了EMC方法的最优设置。
