Hacking Reinforcement Learning

Explore innovative techniques for generating high-quality datasets in reinforcement learning environments through a conference talk delivered at EuroPython 2018. Dive into a new family of planning algorithms that offer superior performance and efficiency compared to existing alternatives. Learn about the four phases of hacking reinforcement learning: information gathering, finding attack vectors, exploiting vulnerabilities, and measuring results. Witness practical demonstrations of sampling superhuman-level game data in real-time using only a laptop. Gain insights into the fundamental concepts behind FractalAI theory and its applications in reinforcement learning. Engage with graphical examples, videos, and a Jupyter notebook to enhance understanding of these cutting-edge techniques.

Guillem Duran - Hacking Reinforcement Learning