# A Survey of Learning in Multiagent Environments: Dealing with Non-Stationarity

**Authors:** Pablo Hernandez-Leal, Michael Kaisers, Tim Baarslag, Enrique Munoz de Cote  
**arXiv:** [1707.09183](https://arxiv.org/abs/1707.09183)  
**Date:** 2017-07-28  
**Summary page:** https://www.summarizepaper.com/en/paper/1707.09183v1/

## Key points

- The paper addresses the challenge of learning a best response to other agents' adaptive strategies in multiagent learning
- It provides an overview of research using game theory, reinforcement learning, and multi-armed bandits to tackle this issue
- The authors propose a taxonomy for classifying algorithms based on their coping mechanisms
- Strengths and limitations of these algorithms are discussed
- Illustrative variations in one domain are presented
- Promising avenues for future research are identified

## Summary

The paper titled "A Survey of Learning in Multiagent Environments: Dealing with Non-Stationarity" by Pablo Hernandez-Leal, Michael Kaisers, Tim Baarslag, and Enrique Munoz de Cote addresses the key challenge in multiagent learning of learning a best response to the behavior of other agents who may adapt their strategy, causing non-stationarity. The authors provide a comprehensive overview of research tackling this issue using game theory, reinforcement learning, and multi-armed bandits. They propose a taxonomy for classifying algorithms based on their coping mechanisms and discuss their strengths and limitations. The paper also presents illustrative variations in one domain and identifies promising avenues for future research. In summary, this paper offers a detailed analysis of how different algorithms address opponent-induced non-stationarity in multiagent learning.