- Abstract: The ability of agents to learn to communicate by interaction has been studied through emergent communication tasks. Inspired by previous work in this domain, we extend the referential game setup to a population of spatially distributed agents. In such a setting, our experiments reveal that multiple languages can emerge in the population and some agents develop multilingual traits. Further, an action-advising framework is proposed for improving sample efficiency in the learning process.
- TL;DR: The paper demonstrates that multilingual traits can emerge in agents playing population based referential games.
- Keywords: Multi-Agent, Reinforcement Learning, Referential Games, Emergent Language