In this lecture, we will cover what was totally lacking in the last editions of AI4Games: Minimax-based family of algorithms. We will go back to the 80s to discuss enhancements that are still state-of-the-art and variants of classic algorithms like PNS or B*, yet also discuss Minimax conjunctions with Neural Networks (e.g. UBFM).
Additionally, in the context of "practical" (or rather "competitive") game AI, we will spend some lectures on proper optimization techniques.
The lab exercises will mainly require implementing various agents aimed at solving small-to-medium-sized problems. Finally, group projects will be an opportunity to develop something bigger.
I will try not to put ~~any~~ too much MCTS-related things this time.
Course subject is inspired by "Intelligent Search & Games" course by prof. Mark Winands, Maastricht University.
Warning: the course is meant to be hardcore. If you will survive until the end, you should pass.
