Overview

Learning is the way,
We reach for higher being,
How shall they now fare?

Machine learning is the ability for artificial intelligence to learn and improve from experience without being reprogrammed. Its’ use in Procedural content generation (PCG) is to use models trained on existing content [1]

How does machine learning for PCG works

Machine learning for PCG works by using techniques like markov chain generation and non-negative matrix. While these methods do learn from existing levels they are unable to change common design decisions like length of a level and difficulty[2]

Another method is to use the kernel methods such as gaussian processes which is popular in approaches for nonlinear regression and classification, a major issue with gaussian processes is that without human intervention they are unable to perform pattern discovery and extrapolation[3]

References

[1] Georgios N. Yannakakis and Julian Togelius. Artificial Intelligence and Games. 2018.
[2] Summerville, Adam J., et al. "The learning of zelda: Data-driven learning of level topology." Proceedings of the FDG workshop on Procedural Content Generation in Games. 2015.
[3] Wilson, Andrew G., et al. "Fast kernel learning for multidimensional pattern extrapolation." Advances in Neural Information Processing Systems. 2014.