A modern Bayesian look at the multi‐armed bandit

A multi‐armed bandit is an experiment with the goal of accumulating rewards from a payoff distribution with unknown parameters that are to be learned sequentially. This article describes a heuristic for managing multi‐armed bandits called randomized probability matching, which randomly allocates observations to arms according the Bayesian posterior probability that each arm is optimal. Advances in Bayesian computation have made randomized probability matching easy to apply to virtually any payoff distribution. This flexibility frees the experimenter to work with payoff distributions that correspond to certain classical experimental designs that have the potential to outperform methods that are ‘optimal’ in simpler contexts. I summarize the relationships between randomized probability matching and several related heuristics that have been used in the reinforcement learning literature. Copyright © 2010 John Wiley & Sons, Ltd.     

常利率下复合泊松-更新风险模型的破产问题

本文研究了常数利率下, 保费收入为复合Poisson 过程, 理赔到达过程为一般更新过程的风险模型. 利用离散化的方法, 获得了该风险模型的破产概率、破产时余额分布及破产前瞬间余额分布的级数展开式, 推广了文[1] 和文[2] 中的相关结果.     

折现率离散时间风险模型下最大赤字问题

在引入折现率的条件下研究离散时间风险模型,运用递推方法和全概率公式,得到了破产前盈余,破产后赤字以及它们的联合分布所满足的微分积分方程,作为推论得到了破产概率所满足的微积分方程并得出结论.     

离散时多元风险模型的破产概率

本文研究了离散时多元风险模型的破产概率问题.利用经典大偏差的方法,获得了有限水平的破产概率,推广了离散时一元风险模型的相应结论.     

一类相依两险种风险模型的分类破产概率

本文考虑文[1]中引入的一类索赔达到计数过程相关的两险种风险模型.利用更新方法,获得了该风险模型的分类破产概率的渐进结果,并给出了指数索赔情形下分类破产概率的表达式,从而改进了文[1]中的相关结果.     

常利率复合二项双险种风险模型的研究

提出了含利率因素的复合二项双险种风险模型,并在有关假设的基础上,给出了此模型下保险公司稳定经营的必要条件;证明了索赔时刻的盈余过程是一马氏过程和调节系数的存在性,并采用递归方法得到了模型的破产概率的上界估计.     

Cramér-Lundberg型下保险公司的最优投资和混合再保

假定保险公司既可以投资在风险资产上,同时又允许混合再保险.用经典的Cramér-Lundberg模型来近似保险公司的盈余过程,考虑了在破产概率最小限制下保险公司的最优投资和再保策略满足的HJB方程,证明了解的存在性和最优性,并对最优策略下的破产概率进行了近似估计.     

Using phase-type models to cost stroke patient care across health,social and community services

Stroke disease places a heavy burden on society, incurring long periods of time in hospital and community care, and associated costs. Also stroke is a highly complex disease with diverse outcomes and multiple strategies for therapy and care. Previously a modeling framework has been developed which clusters patients into classes with respect to their length of stay (LOS) in hospital. Phase-type models were then used to describe patient flows for each cluster. Also multiple outcomes, such as discharge to normal residence, nursing home, or death can be permitted. We here add costs to this model and obtain the Moment Generating Function for the total cost of a system consisting of multiple transient phase-type classes with multiple absorbing states. This system represents different classes of patients in different hospital and community services states. Based on stroke patients’ data from the Belfast City Hospital, various scenarios are explored with a focus on comparing the cost of thrombolysis treatment under different regimes. The overall modeling framework characterizes the behavior of stroke patient populations, with a focus on integrated system-wide costing and planning, encompassing hospital and community services. Within this general framework we have developed models which take account of patient heterogeneity and multiple care options. Such complex strategies depend crucially on developing a deep engagement with the health care professionals and underpinning the models with detailed patient-specific data.     

抽样调查的设计与分析(Ⅰ)

本文讨论抽样调查实践中有关设计和数据分析的若干问题．第一部分主要涉及设计问题．