Throwback Thursday: Old School Optimization Using Newfangled Machine Learning

In the latest industry trend, we're seeing...a renaissance of century old optimization techniques. Artificial intelligence systems can be great, but they come with uncertainty in their output and the nagging feeling that they could just be making s*** up. Old school linear programming and similar optimization systems offer a breath of fresh air to the skeptics in the room. They not only produce optimal solutions, but can produce proof their solutions are optimal.

In a hybrid ML/optimization system, you use machine learning models to forecast parameters that have uncertainty and then use optimization software to make the best possible decision given your forecasts. I'll talk about why I'm such a big fan of these hybrid systems, and how they are shockingly easy to set up using open source software.