Instructors
Carlo Mannino and Torkel Haufmann.
About
Mathematical optimization and learning are strongly intertwined. On one hand, optimization algorithms are the indispensable building blocks of learning methods. On the other hand, learning strategies are increasingly exploited to enhance classical optimization algorithms. These are the two main legs of AI systems: learning allows us to describe and predict data (resp. descriptive and predictive analytics), whereas optimization algorithms use the processed data to make the best possible decisions out of it (prescriptive analytics).
This course will focus on optimization and its application in learning with the intention of introducing students to mathematical optimization and provide them a coherent basis for further study. We will thus touch upon linear and mixed integer programming, combinatorial optimization, flows and minimum cuts, convexity theory, and a pinch of complexity theory. The methods will then be exploited in learning and classification models, such as neural networks and classification trees. We will look at other practical applications throughout.