In 2013, a small company in London called DeepMind uploaded their pioneering paper “Playing Atari with Deep Reinforcement Learning” to Arxiv. In this paper they demonstrated how a computer learned to play Atari 2600 video games by observing just the screen pixels and receiving a reward when the game score increased. The result was remarkable, because the same model architecture, without any change, was used to learn how to win in seven different games, and in three of them the algorithm performed even better than a human!
No wonder DeepMind was immediately bought by Google and has been on the forefront of deep learning research ever since. It has been hailed since then as the first step towards general artificial intelligence – an AI that can survive in a variety of environments.
The roadmap of this seminar is:
– What are the main challenges in reinforcement learning?
– How to formalize reinforcement learning in mathematical terms?
– How do we form long-term strategies?
– How can we estimate or approximate the future reward?
– What if our state space is too big? (Here the answer is simple: deep learning!)
– What are the main deep RL algorithms?
– What performances can these algorithms obtain on classical problems?
– What are the main issues of applying such algorithms?
– How to program such algorithms and validate them on common problems with Ipseity?