Policy Iteration for Learning an Exercise Policy for American Options

Learning an Exercise Policy for American Options is a trending topic that has a lot of researches. Policy Iteration is one of the areas of research on this topic. As part of this work, a general review of the topics related to the execution of American options and Monte Carlo methods in relation to finance was made. After that, the parametric method and the random tree method were reviewed, those methods are quite common in optimizing the exercise of American options. Also, the modern methods which are “Deep Optimal Stopping” and “Optimal stopping via reinforced regression” are analyzed in more detail and a general comparison of the methods and their features are made. Further, the code has been implemented and results are obtained on the following options for exercising American options: - Bermudan max-call option - - Symmetric case - - Asymmetric case - Callable multi barrier reverse convertibles (CMBRC) With respect to the obtained data comparison of the effectiveness between “Deep Optimal Stopping” and “Optimal stopping via reinforced regression” methods is made, also the options for improving the efficiency of algorithms and further development of the work were proposed.