program

WaveFunctionCollapse (WFC) is the popular name for a family of algorithms that combine machine learning and constraint solving for procedural generation on discrete graphs. WFC was developed and popularized in 2016 by Maxim Gumin, building on past research on texture synthesis and model synthesis. While the reference implementation of WFC is written in C# and uses the AC-4 algorithm, WFC has since been implemented in dozens of programming languages and environments, with variations in both the constraint solving and machine learning components.

WFC is a versatile algorithm, and has been implemented in environments spanning from a general-purpose constraint solver to a tiny retrocomputing fantasy console. Both the machine learning and constraint solving aspects of the system can be varied, which has made for a wide scope for potential research. Many recent publications have been exploring mixed-initiative co-creation applications, combining WFC with neural networks, and other applications. Further, as an algorithm that produces understandable results while being within the technical capabilities of individual developers, it has seen some popularity among artists and game developers.

In this tutorial, we will briefly demonstrate how to implement WFC, how to apply it to a few standard problems, and how to extend it to novel or underexplored possible areas of application.

Statistical Forward Planning (SFP) is a group of robust and general AI techniques that use a simulation model to adaptively search for effective sequences of actions in various games and other problems characterised as Markov Decision Processes (MDPs, including Partially Observable – POMDPs).

SFP methods can operate without the need for prior training and can handle complex and dynamic environments. This tutorial will provide a tour of SFP from basics to finer points, complete with pointers to Python code (e.g. in OpenSpiel and other repositories). We will cover a number of powerful SFP algorithms including Monte Carlo Tree Search (MCTS), Rolling Horizon Evolutionary Algorithm (RHEA) and Monte Carlo Graph Search (MCGS), as well as handling partial observability with Information Set MCTS. We’ll also cover:

• the relationship between SFP algorithms and Counterfactual Regret Minimisation (CFR)
• incorporating policy and value functions (similar to AlphaZero) 
• efficient exploration functions for flat reward landscapes
• handling combinatorial action spaces

Demonstrations will show these algorithms can play a variety of video games surprisingly well and provide insights into their working principles and behaviours. The tutorial will be suitable for those with no experience of SFP, but we also expect MCTS veterans to gain some fresh insights. We will conclude with a discussion of some of the most exciting challenges in the area.

Simon Lucas – Queen Mary University of London
James Goodman – Queen Mary University of London
Marko Tot – Queen Mary University of London

 

Speakers Bio

Simon is a full professor in the School of Electronic Engineering and Computer Science at Queen Mary University of London (QMUL) where he leads the Game AI Research Group. He was previously Head of School of EECS at QMUL, and prior to that, Head of School of Computer Science and Electronic Engineering at the University of Essex. He recently spent two years as a research scientist / software engineer in the Simulation-Based Testing team at Meta, applying simulation-based AI to automated testing.

Simon was the founding Editor-in-Chief of the IEEE Transactions on Games and co-founded the IEEE Conference on Games, was VP-Education for the IEEE Computational Intelligence Society and has served in many conference chair roles. His current research is focused on simulation-based AI and sample efficient optimisation.

James Goodman is in his final year of PhD on statistical forward planing in multiplayer board games. He is currently focusing on working with tabletop games designers to use AI techniques in TAG to test their games during the design phase. He indecisively has first degrees in Chemistry, Mathematics and History, and an MSc in Computational Statistics and Machine Learning from UCL.

Marko Tot is a PhD student at Queen Mary University of London and a part of the GameAI research group. He is the recipient of the Microsoft Research PhD Studentship in 2020, working on combining Learned Forward Models with Statistical Forward Planning methods, with a current focus on state representation in Monte Carlo Graph Search.

Procedural Content Generation via Knowledge Transformation (PCG-KT) is a framework covered in a newly-accepted Transactions on Games paper for organizing approaches that transform (typically learned) knowledge about a game to approximate another game or one that does not exist. Conceptual Blending, Domain Adaptation, and Transfer Learning are all existing approaches to this task. In this tutorial, we will overview PCG-KT and give practical, hands on experience through provided code of conceptual blending, domain adaptation, and transfer learning approaches to it. We will end with the opportunity for participants to propose and discuss possible future applications or research projects in this area.

Matthew Guzdial – University of Alberta

 

Speakers Bio

Matthew Guzdial is an Assistant Professor in the Computing Science department of the University of Alberta and a Canada CIFAR AI Chair at the Alberta Machine Intelligence Institute (Amii). His work has focused on the intersection of procedural content generation and machine learning, with features in the BBC, WIRED, Popular Science, and Time.