Stochastic formulation of sampling dynamics in generalized ensemble methods.
Stochastic formulation of sampling dynamics in generalized ensemble methods.
复制标题
广义集成方法中采样动力学的随机公式。
DOI：
--
复制
发表时间：
2004
期刊：
影响因子：
--
通讯作者：
Haruki Nakamura
中科院分区：
文献类型：
--
作者：
Jae Gil Kim;Y. Fukunishi;A. Kidera;Haruki Nakamura
在扩展系综形式下，研究了多规范系综抽样过程与模拟回火的基本关系。模拟回火被识别为多个典型的采样与广义的权重确定的温度权重的拉普拉斯变换。这两种采样方法的特征动力学已被验证的随机制定的采样过程。我们的研究给出了权值在能量和温度空间中实现均匀采样的充分必要条件。基于随机模型，鲁棒力偏置迭代计划已被提出，允许自动确定均匀的采样权重。
The fundamental relations of the sampling process in the multicanonical ensemble and simulated tempering have been studied in the expanded ensembles formalism. The simulated tempering is identified as the multicanonical sampling with the generalized weight determined by a Laplace transform of the temperature weight. The characteristic dynamics of both sampling methods has been verified in the stochastic formulation of the sampling process. Our study gives a necessary and sufficient condition for the weights to realize a uniform sampling in the energy and temperature spaces. Based on the stochastic model, robust force biased iteration schemes have been proposed to allow automatic determinations of uniform sampling weights.