Robust Optimization using Machine Learning for Uncertainty Sets
This work addresses the challenge of making decisions robust to uncertainty in complex, high-dimensional data for optimization practitioners, representing an incremental improvement over classical methods.
The paper tackles the problem of designing robust optimization policies by learning uncertainty sets from past data, rather than using user-chosen or simplistic sets, and demonstrates that this approach provides probabilistic guarantees on robustness.
Our goal is to build robust optimization problems for making decisions based on complex data from the past. In robust optimization (RO) generally, the goal is to create a policy for decision-making that is robust to our uncertainty about the future. In particular, we want our policy to best handle the the worst possible situation that could arise, out of an uncertainty set of possible situations. Classically, the uncertainty set is simply chosen by the user, or it might be estimated in overly simplistic ways with strong assumptions; whereas in this work, we learn the uncertainty set from data collected in the past. The past data are drawn randomly from an (unknown) possibly complicated high-dimensional distribution. We propose a new uncertainty set design and show how tools from statistical learning theory can be employed to provide probabilistic guarantees on the robustness of the policy.