Clustering-based Tile Embedding: A General Representation for Level Design
In this EXAG 2022 presentation, Mrunal Jadhav discusses how there has been significant research interest in general Procedural Level Generation via Machine Learning (PLGML), applying ML techniques to automated level generation. One recent trend is in the direction of learning representations for level design via embeddings, such as tile embeddings. Tile Embeddings are a continuous vector representation of levels unifying its visual, contextual and behavioral information. However, the original tile embedding struggled to generate levels with skewed tile distributions. For instance, Super Mario Bros. (SMB) wherein a majority of tiles represent the sky or background. To remedy this, they propose a new tile embedding representation and a novel, two-step level generation to leverage this representation. They evaluate the performance of their approach in generating levels for seen and unseen games with skewed tile distributions and compare their performance against the original tile embedding-based generator.