Team Members
Team Awards
The series of papers address a fundamental challenge in optimization under uncertainty: that the distribution of the uncertain problem parameters, which is needed to compute the expected value of the objective function, is unknown. In practice, one has access to a set of training samples from this distribution. In this case, a natural goal is to find a procedure that transforms the training data to a hopefully near-optimal decision and a prediction of its expected cost. The papers construct a data-driven approach to decisions by solving a distributionally robust optimization problem over a Wasserstein ball. These contributions are not only foundational but they have also paved the way for a new perspective on popular methods in statistics and machine learning, and as well as applications.