Sutton and Barto’s standard textbook on Reinforcement Learning covers how state feature vectors may be constructed for linear state value functions. However, there is little explanation of the extension to state-action feature vectors. In this post I aim to fill this gap.
The Papyrus scripting language of The Elder Scrolls V: Skyrim is somewhat notorious in the game’s large community of third-party mod developers and users for performance-related issues (though there are also plenty of misconceptions, and it’s not necessarily as bad as many people think). There are already many relat...
Ludii is a new general game-playing (GGP) system, of which an early pre-release version has been published today. The system comes with a wide range of built-in games, and may be interesting for game-playing enthusiasts to simply play games! Ludii’s new game description language tends to lead to game descriptions th...
This post provides an informal description of my MaastCTS2 agent for General Video Game Playing. It participated in the Single-Player and Two-Player tracks of the GVG-AI Competition in 2016. The results can be found here. MaastCTS2 ended up winning the Single-Player track, and was the runner-up in the Two-Player tra...