A Brief Introduction to Reinforcement Learning
- Wednesday 8 February 2023
Niels Bohrweg 1
2333 CA Leiden
Sequential decision-making is one of the key challenges in artificial intelligence research. One approach to this problem is reinforcement learning (RL), where we attempt to learn behaviour through trial and error. This approach has shown empirical success in recent years, for example defeating the human world champion in the game of Go. The present talk aims to provide a gentle introduction to reinforcement learning, focusing on the key conceptual topics. We discuss the problem setting, its potential benefits, as well as the challenges that arise. We then cover common solution approaches, for example discussing the exploration-exploitation trade-off, the credit assignment problem, deep reinforcement learning, and model-based reinforcement learning. Altogether, the talk aims to provide a starting point for further education and/or research in reinforcement learning.