PhD School: Reinforcement Learning for Operations Research
The goal of this PhD school is to familiarize doctoral students in stochastic modelling with reinforcement learning (RL) techniques through a hands-on, hackathon-style workshop.
The school will cover a range of methods (preliminary list), including:
- Stochastic dynamic programming
- Monte Carlo simulation
- Temporal-difference learning
- Value function approximation
- Eligibility traces
- Policy gradient methods
- Bayesian dynamic programming
- Multi-armed bandits
- Deep reinforcement learning
The program includes lectures by specialists, both in person and online, combined with dedicated time for participants to work in small groups on selected problems. Experts will actively engage with the groups through discussions and Q&A sessions.
Technical Program
- Sun 17 Jul: Arrival before dinner; non-technical overview of RL and problem areas
- Mon 18 Jul: Lectures on techniques; discussion; identification of problems; group formation
- Tue 19 Jul: Group implementation; interaction with specialists
- Wed 20 Jul: Lectures; discussion; identification of problems
- Thu 21 Jul: Group implementation; interaction with specialists
- Fri 22 Jul: Lectures; discussion; identification of problems
- Sat 23 Jul: Group implementation; interaction with specialists; presentation of results
- Sun 24 Jul: Evaluation of group work; discussion on the future of RL in OR; departure after lunch
Social Program
Two organized social events will be complemented by several informal activities to encourage interaction among participants.
The detailed program can be found here.