Alpha Go Zero left us with our jaws dropped. The Dota2 and StarCraft2 agents did so even more. And watching the output of companies like DeepMind leaves us stunned in awe. But how do all these systems work? What is this „deep reinforcement learning“ magic?

In this session we will first learn the core ideas of reinforcement learning – a bit of math (not too much, promised!), algorithms, learning strategies and more. We will also see how to implement a reinforcement learning agent in practice. Then we will extend the approach to deep reinforcement learning by adding deep neural networks. To complete the picture, we will have a quick look at the current software and service ecosystem.

After the session, we will not have built the next Alpha Go Zero, but you’ll have an idea how you could …