Michael Wunder
Abstract: In most real-world strategic situations, the basic rules and payoffs of the game are changing all the time. However, game theory is based on the idea that payoffs are fixed, which simplifies the task of finding equilibria. The more recent iterations of the Lemonade Stand Game demonstrate dynamic environments, where experience gained in a prior interaction is not directly applicable to the current one. The task at hand for multiagent researchers is how to use machine learning to predict behaviors of anonymous agents in novel environments and construct successful strategies in response.