Efficient and Scalable LP-based Multi-Stage Decision Making in R

Multi-stage decision-making problems widely appear and pose unique challenges in various real-world applications, e.g., robot control, game play, and healthcare. Consider the clinical trial design as an example. Optimizing the clinical trial design is essential to decrease the cost of drug development, as the average costs are up to 52.9 million US dollars for a Phase 3 trial [1]. The current approach formulates multi-stage decision-making as a general Linear Programming (LP) problem, which however, is computationally expensive especially in large-scale settings. In this project, we aim to develop new customized algorithms and an R package for multi-stage decision-making problems with three key features: 1) It provides a highly efficient solver to tackle a large and important class of sparse LP problems; 2) It provides a solution for multi-stage decision-making problems with Bayes risk constraints; 3) It provides additional functions such as visualization of the optimal decision maps.