Multi-objective sequence learning for Chemistry and Computer games

Mike Preuss

Leiden University, The Netherlands


Monte Carlo Tree Search is a popular method for dealing with complex, highly branched tree search problems that represent sequences, be it steps to win a game or reactions to make a molecule. However, nearly all available algorithm variants deal with one objective only. But what if we have multiple objectives? Up to now, there are very few methods for this and I report on our attempts to use these methods for retrosynthesis (find ways how to make a specific target molecule according to several criteria) and also in game AI. It seems clear that this is a research area with high potential but little activity as of now.

Short Bio

Mike Preuss is Associate Professor at LIACS, the computer science institute of Universiteit Leiden in the Netherlands. Previously, he was with ERCIS (the information systems institute of WWU Muenster, Germany), and before with the Chair of Algorithm Engineering at TU Dortmund, Germany, where he received his PhD in 2013 for his work in evolutionary algorithms for multimodal optimization. His current research focuses on Game AI algorithms as reinforcement learning and generative AI methods and their application to real world problems as procedural content generation, social media computing, and hard application problems in Chemistry, as retrosynthesis.