This is a talk given by Prof Adam Elmachtoub as part of the Online Seminar Series Machine Learning NeEDS Mathematical Optimization, https://congreso.us.es/mlneedsmo/, branding the role of Operations Research in Artificial Intelligence with the support of EURO, https://www.euro-online.org.
A standard problem for any analytics researcher or practitioner is to first predict unknown parameters using a machine learning model, and then plug in those parameters into a downstream optimization problem to make a decision. We coin this commonly used practice as predict-then-optimize. In this talk, we discuss a general framework called Smart Predict-then-Optimize (SPO) that integrates the machine learning and optimization components for the setting of contextual linear optimization. We show that our SPO framework lends itself to computationally tractable formulations for training machine learning models that yield improved decisions over natural benchmarks. We also provide theoretical guarantees on consistency and generalization to help explain the performance. This is joint work with Paul Grigas, Ambuj Tewari, Ryan McNellis, Jason Liang, and Othman El Balghiti.
Link to the talk: https://eu.bbcollab.com/guest/40a20d4cc309494fa78f3d0b88102365
The organizers of the Online Seminar Series Machine Learning NeEDS Mathematical Optimization
Emilio Carrizosa, IMUS-Instituto de Matemáticas de la Universidad de Sevilla
Nuria Gómez-Vargas, IMUS-Instituto de Matemáticas de la Universidad de Sevilla
Thomas Halskov, Copenhagen Business School
Dolores Romero Morales, Copenhagen Business School