Reinforcement Learning (RL), the field of sequential decision making, has evolved significantly within the last few years, achieving super-human performance in solving complex board games, 2D Atari and 3D games (Doom, Quake, StarCraft). But this is not just about games, this is about solving arbitrary problems with truly general algorithms. Fully dedicated to this pursuit, there is a huge research community involved with the ultimate goal of creating an “artificial general intelligence”.
[Related Article: Reinforcement Learning vs. Differentiable Programming]
This talk will give an overview of the main breakthroughs, paradigms, formulations, and obstacles within RL to quickly get you up to speed with where the field currently is and where it is going.