Controlled Diffusion Model for Optimal Dividend Payment Under the Reinforcement Learning Formulation

  • A+

:2021-06-21 14:30
:腾讯会议ID:291 554 938(无设置密码)


时  间:621日下午14:30

地  点:腾讯会议ID291 554 938(无设置密码)


In this paper we investigate the optimal dividend problem under the reinforcement learning and continuous time entropy-regularized formulation. More precisely, we try to give the optimal dividend control when the state process is governed by a diffusion process under the reinforcement learning formulation. We use the viscosity solution method to prove existence and uniqueness of smooth solution of the Hamilton-Jacobi-Bellman equation and that the solution is bounded and concave. We show that the optimal control is no longer a threshold control, but a truncated distribution which has a ``gradual'' form. Then, we provide an example of this optimal control. Finally, we show that the condition of the verification theorem holds and the entropy-regularized stochastic control problem of dividend payment degenerates to the classical form.


柏立华,理学博士,南开大学数学科学学院教授、博士生导师。入选教育部新世纪优秀人才支持计划(2013)、天津市“131”创新型人才培养工程第二层次人选(2014)、天津市青年拔尖人才支持计划(2015)、天津市创新人才推进计划青年科技优秀人才(2017)。获全国优秀博士学位论文提名奖(2012)、天津市数学会青年学术奖一等奖(2017)。其主要研究方向包括随机过程、随机控制、精算数学、金融数学等。目前已经在AAPSICONSPABernoulliJOTAIMEScience China (A)SAJ等主流期刊发表论文20余篇。作为项目负责人,主持国家自然基金面上项目1项,青年基金1项,天津市青年拔尖人才支持计划1项,南开大学百名青年学科带头人培养计划”1项。