Название: Procedural Content Generation Via Machine Learning: An Overview, 2nd Edition
Автор: Matthew Guzdial, Sam Snodgrass, Adam Summerville
Издательство: Springer
Год: 2025
Страниц: 300
Язык: английский
Формат: pdf (true), epub
Размер: 38.5 MB

This second edition updates and expands upon the first beginner-focused guide to Procedural Content Generation via Machine Learning (PCGML), which is the use of computers to generate new types of content for video games (game levels, quests, characters, etc.) by learning from existing content. The authors survey current and future approaches to generating video game content and illustrate the major impact that PCGML has had on video games industry. In order to provide the most up-to-date information, this new edition incorporates the last two years of research and advancements in this rapidly developing area. The book guides readers on how best to set up a PCGML project and identify open problems appropriate for a research project or thesis. The authors discuss the practical and ethical considerations for PCGML projects and demonstrate how to avoid the common pitfalls. This second edition also introduces a new chapter on Generative AI, which covers the benefits, risks, and methods for applying pre-trained transformers to PCG problems.

Procedural Content Generation (PCG) refers to the algorithmic generation of game content. By game content, we mean the various component parts of a game, including scripts or game code, levels or maps, character sprites or 3D models, animations, music, sound effects, and so on. By algorithmic generation we indicate some process or set of rules rather than typical human creation. Most often “algorithmic” indicates the use of computer code, and that’s the way we’ll discuss it in this book, but it could involve any set process or rules, such as generation via cards or dice. PCG can look like the landscapes of Minecraft, the conversations and dungeons of Hades, or almost everything present in No Man’s Sky.

Who Is This Book For?
Our hope is that this book is accessible to PCG practitioners, ML practitioners, and anyone interested in these topics. The book can be used as the basis for a class, with every chapter serving as the basis of 1–2 lectures, as an introduction to these topics, or simply as a reference or guide. Our hope is that this book can demystify ML for those on the game design and PCG side of things, and make the benefits of applying ML to PCG clear for those on the ML side of things. For a class, we have written this book to be programming language agnostic, but it will require at least some understanding of coding. We recommend using this book for students at least at the senior undergraudate level, as many of the concepts in the book rely on fairly complex mathematics, though we’ll do our best to express these clearly. Our hope is that the individual chapters can serve as a reference and guide for individuals looking to implement particular PCGML approaches, or for those interested in conducting PCGML research.