Emergent Quantized Communication

TL;DR

This paper introduces a quantization method for emergent multi-agent communication, enabling end-to-end training of discrete messages with better performance than previous continuous or reinforcement learning approaches.

Contribution

It proposes message quantization as a natural and effective way to achieve discrete communication in multi-agent systems, bridging the gap between continuous and discrete messaging.

Findings

Quantization enables end-to-end training of discrete communication.

Quantized communication outperforms reinforcement learning and continuous methods.

The approach offers a unified framework from continuous to discrete messaging.

Abstract

The field of emergent communication aims to understand the characteristics of communication as it emerges from artificial agents solving tasks that require information exchange. Communication with discrete messages is considered a desired characteristic, for both scientific and applied reasons. However, training a multi-agent system with discrete communication is not straightforward, requiring either reinforcement learning algorithms or relaxing the discreteness requirement via a continuous approximation such as the Gumbel-softmax. Both these solutions result in poor performance compared to fully continuous communication. In this work, we propose an alternative approach to achieve discrete communication -- quantization of communicated messages. Using message quantization allows us to train the model end-to-end, achieving superior performance in multiple setups. Moreover, quantization is a natural framework that runs the gamut from continuous to discrete communication. Thus, it sets the ground for a broader view of multi-agent communication in the deep learning era.