Reinforcement Learning and Long-Term Value Prediction

In recent years, the conjunction of Deep Learning and Reinforcement Learning (RL) has achieved great successes in the areas of robotics and computer games (Atari, Go, Starcraft, DOTA). In this talk we will introduce some of the fundamental concepts of RL, and give you pointers for you to play with those even inside your browser. We will then explain how the same framework allows you to think of other sequential decision making processes, and how you can predict the long-term value of your actions. This translates directly into potential real-world applications ranging from recommender systems to complex industrial controls.