模拟退火马氏链的尾事件结构

On structare of the Tail Events of Simulated Annealing Markov Chains

由模拟退火算法所引出的非时齐马氏链问题已经引起概率论学者的极大关注。本文试图按其轨道的渐近性质将状态空间进行分类，区分了常运态和瞬时态，且将常运态划分为一些渐进连通的互不相交的子类，而这些子类对应于尾σ－代数的原子集。

Discussed in this paper is non-homogenous Markov chains generated by the simulated annealing algorithm. According to asymptotic properties of trajectories we classify states under the weak reversibility assumption of Hajek. And we have obtained the classes corresponding to atoms of tail σ-field.