摘要
小概率事件又称不可能事件。由于其发生概率极低,使得用传统的蒙特卡洛方法预测概率比较困难。因此,寻找新的估计方法变得十分必要。本文将交互粒子系统方法运用于复杂系统的小概率事件概率估计中,为小概率事件的估计提供了新的思路。本文首先通过建模将小概率事件转化为阈值问题,并利用马尔科夫链对其进行模拟。其次,本文采用目前比较热门的交互粒子系统方法,对小概率事件进行估计,并分析了筛选度与筛选步长对概率估计的影响,进而进行优化。最后,本文比较了交互粒子系统方法与蒙特卡洛方法的效率和精度。
Rare event has a quite low occurrence probability so that it’s hardly estimated by Monte Carlo method.It is necessary to provide new algorithms.In this contribute,interacting particle systems(IPS)is applied,in which,trajectories with more possibility to reach target event are multiplied and the others are killed.The rare event is firstly modeled by threshold exceedance problem and then simulated by Markov process.Moreover,we noticed that the performance of IPS is related to two parameters,selection degree and selection distance.An optimization will also be studied by adjusting the two parameters.Finally,we compare this method with Monte Carlo method about the efficiency and accuracy.
出处
《科技创新导报》
2017年第20期14-15,共2页
Science and Technology Innovation Herald
关键词
小概率事件
蒙特卡洛方法
交互粒子系统
Rare event probability
Monte Carlo
Interacting particle systems
