15 October 2019, ILLC Lecture, Daan Bloembergen

Abstract:

Many real-world problems require in their solution the cooperation between multiple autonomous decision makers. How exactly to achieve this may not be straightforward from the start, since long-term social welfare often needs to be balanced against short-term individual gain. In this talk I will give a brief overview of those aspects of multi-agent learning that make cooperation a challenging task. Then, I will provide examples from my own work that aim to elicit cooperative solutions, ranging from individual (reinforcement) learning approaches to population-based analysis.