How to teach space invaders to your computer
A very brief introduction to reinforcement learning theory followed by a hands on section in which it is demonstrated how to train an algorithm to play space invaders. After this talk you should now what reinforcement learning is and where to dig deeper if you like it.
Tags: Deep Learning & Artificial Intelligence, Jupyter, Python, Science
Scheduled on thursday 16:00 in room media
David Wölfle studied mechanical engineering in Karlsruhe. During his master studies he focused on wind energy technology and researched data sources for wind resource estimation. After graduating as Master of Science from Flensburg University of applied Applied Sciences in 2015, David worked as a R&D Scientist scientist at EWC Weather Consult GmbH (now UBIMET GmbH), where he designed and implemented software components for the estimation and predication of renewable energy power production. In 2016 David has been promoted to a team manager at EWC Weather Consult where he was responsible for the software engineering within the product development and as well as the design and execution of the project management. Besides these duties, he also developed innovative methods for estimating power production of airborne wind energy converters, using high high-resolution meteorological data and machine learning methods.
Since early 2018 David works as a research scientist at FZI Research Center for Information Technology in the field of smart energy. His research focuses thereby on self-learning energy management systems using reinforcement learning techniques.
First things first: playing good old Atari games might be cool but why should I write a program for doing it? Well teaching a computer to play a game means teaching it to develop strategies and use foresight planning to solve a certain problem. The tools you gather while solving i.e. space invaders are the same you may use to solve any problem which requires a sequential set of decisions in order to find an optimal solution to some problem, like i.e. controlling a robot that collects garbage. Furthermore, there is a lot of scientific research on reinforcement learning that focuses on solving Atari games which makes it a good starting point, as large amounts of publications and open source code already exists.
What to expect from this talk? At first there will be a very short introduction to reinforcement learning theory, just the very basics, common applications and some references for further reading. Next points are, how to run Atari games from inside python for a learning task (with OpenAI's gym), and where to find an algorithm for the actual learning problem. Finally it will be shown how to build it all together in a jupyter notebook and let the algorithm play the game. Et voilà that's your computer beating you in space invaders.