OR and Uncertainty: A Unified Framework POSTPONED

Sequential decision analytics is the study of making decisions over time under uncertainty. These are the problems at the heart of the transition to automated systems that are being increasingly used in e-commerce, digital transportation, supply chain management, process automation, and robotics.

These problems have been studied by over 15 different communities under names that include dynamic programming (including approximate dynamic programming and reinforcement learning), stochastic programming, stochastic search, stochastic control, and simulation optimization, as well as multiarmed bandit problems and active learning. Each community has evolved its own modeling style and family of algorithms designed for certain classes of applications. As each community has evolved to address a broader range of problems, there has been a consistent pattern of rediscovery of tools that sometimes differ in name only, or modest implementation details.

I will represent all of these communities using a single, canonical framework that mirrors the widely used modeling style from deterministic math programming or optimal control. The key difference when introducing uncertainty is the need to optimize over policies. I will show that all the solution strategies suggested by the research literature, in addition to some that are widely used in practice, can be organized into four fundamental classes. One of these classes, which we call “parametric cost function approximations,” is widely used in practice, but has been largely overlooked by the academic community. These ideas will be illustrated using a variety of applications.

Registration

Deadline for registration is 15 April please click here for more information