Emergent Communication: Generalization and Overfitting in Lewis Games

Author name not available (Why is that?)

Publication date: 30 September 2022

Abstract: Lewis signaling games are a class of simple communication games for simulating the emergence of language. In these games, two agents must agree on a communication protocol in order to solve a cooperative task. Previous work has shown that agents trained to play this game with reinforcement learning tend to develop languages that display undesirable properties from a linguistic point of view (lack of generalization, lack of compositionality, etc). In this paper, we aim to provide better understanding of this phenomenon by analytically studying the learning problem in Lewis games. As a core contribution, we demonstrate that the standard objective in Lewis games can be decomposed in two components: a co-adaptation loss and an information loss. This decomposition enables us to surface two potential sources of overfitting, which we show may undermine the emergence of a structured communication protocol. In particular, when we control for overfitting on the co-adaptation loss, we recover desired properties in the emergent languages: they are more compositional and generalize better.

Has companion code repository: https://github.com/mathieurita/population

This page was built for publication: Emergent Communication: Generalization and Overfitting in Lewis Games

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6412450)