论文标题
用二进制决策树有效地模拟离散状态模型
Efficiently simulating discrete-state models with binary decision trees
论文作者
论文摘要
随机模拟算法（SSA）被广泛用于数值研究随机，离散状态模型的性质。 Gillespie Direct方法是杰出的SSA，广泛用于生成所谓的基于代理或基于个体的模型的样本路径。但是，吉莱斯皮直接方法的简单性通常会使大规模模型进行详细分析，这通常使其不切实际。在这项工作中，我们仔细修改了Gillespie Direct方法，以便它使用自定义的二进制决策树来追踪感兴趣模型的样本路径。我们表明，可以构建决策树以利用所选模型的特定特征。具体而言，将模型的事件放置在决策树的精心选择的叶子中，以最大程度地减少使树保持最新所需的工作。我们意识到的计算效率可以提供调查大规模，离散状态模型所必需的设备，这些设备原本是棘手的。提出了两项案例研究以证明该方法的效率。
Stochastic simulation algorithms (SSAs) are widely used to numerically investigate the properties of stochastic, discrete-state models. The Gillespie Direct Method is the pre-eminent SSA, and is widely used to generate sample paths of so-called agent-based or individual-based models. However, the simplicity of the Gillespie Direct Method often renders it impractical where large-scale models are to be analysed in detail. In this work, we carefully modify the Gillespie Direct Method so that it uses a customised binary decision tree to trace out sample paths of the model of interest. We show that a decision tree can be constructed to exploit the specific features of the chosen model. Specifically, the events that underpin the model are placed in carefully-chosen leaves of the decision tree in order to minimise the work required to keep the tree up-to-date. The computational efficencies that we realise can provide the apparatus necessary for the investigation of large-scale, discrete-state models that would otherwise be intractable. Two case studies are presented to demonstrate the efficiency of the method.
