Machine Learning for Mathematical Optimization / Fall 2021
Updates
- New Lecture is up: Other settings where learning aids optimization [slides] [video]
- New Lecture is up: Learning Theory [slides] [video]
- New Lecture is up: Graph Neural Networks [slides] [video]
- New Lecture is up: Reinforcement Learning for Algorithms [slides] [video]
- New Lecture is up: Project talk + Learning to Branch [slides] [video]
- New Lecture is up: Learning in Exact Solvers [slides] [video]
- New Lecture is up: Algorithm Configuration [slides] [video]
Course Description
This course introduces automated machine learning approaches for improving optimization algorithms in the presence of a historical dataset or a generator of problem instances from a domain of interest. Topics include automated algorithm configuration, modeling iterative heuristics in the reinforcement learning framework, deep neural networks for modeling combinatorial optimization problems, guiding exact solvers with learned search strategies, learning-theoretic guarantees, and benchmarking/computational considerations. The focus will be on discrete optimization in the integer programming framework, both exact and heuristic.
The syllabus is available here.
