Artificial Intelligence (AI) for Computer Games

About

Basics of machine learning (one day + practical component) What is a machine learning model? Differences between supervised, unsupervised and reinforcement learning modes of learning. Focus is on deep learning models, the basics of it will be quickly revised. Recently introduced self-supervised deep learning models will be introduced.

AI agents (two days of lectures + practical component) The course introduces the basics of reinforcement learning and how autonomous systems can be implemented and trained. In the course we learn how to train software agents (such as playing games, dialog systems, etc.) and how to train physical agents such as robots through simulations.

Practice component (two days) During the section, students implement a software agent that can play a computer game by programming (Python). An agent is a statistical model whose parameters are then learned by playing the game. There are many different models and training algorithms, and it is expected that advanced students will be able to try several of these during these two days. During the final day, students’ agents are pitted against each other in a tournament and we will recognize three best performing agents.

Course leader

Ville Hautamäki (ville.hautamaki@uef.fi)

Target group

Master, Doctoral

Course aim

Learn that training an AI agent means that we need to train an agent to make decisions in a sequence. Learn how to formalize this problem and how such problems can be solved. Learn software tools that greatly assist in training an agent. Learn also differences in training a software agent, such as being able to play computer games, and a physical agent, such as a robot.

Fee info