I will use this website to write about my research and maintain a list of my publications and public datasets. I brief introduction to what I do:
At the IT University of Copenhagen I am investigating how deep learning in combination with online evolutionary planning can be used to play strategy games. My focus so far have been on multi-action adversarial games, such as Hero Academy, and the real-time strategy game StarCraft. I am investigating (1) supervised learning methods, where a neural network learn to predict actions of human players, (2) deep reinforcement learning methods, where a neural network is trained from self-play or against other artificial agents, and (3) online evolutionary planning algorithms that can evolve action sequences in games with large branching factors. My publications are listed in the end of this document.
You can see a list of my publications here.