Reinforcement learning applied to operations research: 17.07.-24.07.2022

This PhD school made students in stochastic modelling familiar with several Reinforcement Learning (RL) techniques. The participants applied these concepts to solve Operations Research problems through a hackathon-like workshop. The school contained Lectures from specialists, both at the location and online, and time to work in small groups on various problems.

Participants were divided into groups with respect to their preferred application area: inventory management, queueing systems, and pricing. After a brief introduction to Markovian Decision Processes, the participants had to find appropriate problems to solve with Temporal Difference Learning, Deep Reinforcement Learning, and Multi-Armed Bandits. After learning about different algorithms in tutorials, students implemented these and compared them with each other. Analyzed problems and results are presented and discussed with experts.

The facilities in Castle Gimborn allowed for concentrated work, fruitful discussions and nice get togethers beside the official program. The social program included a walking tour and a dinner in a traditional German Beer Garden. We are looking forward to seeing some of the participants presenting results related to the school in upcoming conferences.