Toward Co-creative Dungeon Generation via Transfer Learning

TL;DR

This paper proposes a transfer learning approach to improve co-creative dungeon generation in games by approximating human-AI interaction data, reducing the need for extensive co-creative training data.

Contribution

It introduces a transfer learning method to adapt co-creative content generation models across different games, specifically for Zelda dungeon room creation.

Findings

Transfer learning effectively adapts models to new game contexts.

Approximate human-AI interaction data can substitute real co-creative training data.

Improved dungeon generation quality demonstrated in experiments.

Abstract

Co-creative Procedural Content Generation via Machine Learning (PCGML) refers to systems where a PCGML agent and a human work together to produce output content. One of the limitations of co-creative PCGML is that it requires co-creative training data for a PCGML agent to learn to interact with humans. However, acquiring this data is a difficult and time-consuming process. In this work, we propose approximating human-AI interaction data and employing transfer learning to adapt learned co-creative knowledge from one game to a different game. We explore this approach for co-creative Zelda dungeon room generation.