Emergent Social Learning via Multi-agent Reinforcement Learning

Kamal K Ndousse, Douglas Eck, Sergey Levine, Natasha Jaques
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:7991-8004, 2021.

Abstract

Social learning is a key component of human and animal intelligence. By taking cues from the behavior of experts in their environment, social learners can acquire sophisticated behavior and rapidly adapt to new circumstances. This paper investigates whether independent reinforcement learning (RL) agents in a multi-agent environment can learn to use social learning to improve their performance. We find that in most circumstances, vanilla model-free RL agents do not use social learning. We analyze the reasons for this deficiency, and show that by imposing constraints on the training environment and introducing a model-based auxiliary loss we are able to obtain generalized social learning policies which enable agents to: i) discover complex skills that are not learned from single-agent training, and ii) adapt online to novel environments by taking cues from experts present in the new environment. In contrast, agents trained with model-free RL or imitation learning generalize poorly and do not succeed in the transfer tasks. By mixing multi-agent and solo training, we can obtain agents that use social learning to gain skills that they can deploy when alone, even out-performing agents trained alone from the start.

Cite this Paper


BibTeX
@InProceedings{pmlr-v139-ndousse21a, title = {Emergent Social Learning via Multi-agent Reinforcement Learning}, author = {Ndousse, Kamal K and Eck, Douglas and Levine, Sergey and Jaques, Natasha}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {7991--8004}, year = {2021}, editor = {Meila, Marina and Zhang, Tong}, volume = {139}, series = {Proceedings of Machine Learning Research}, month = {18--24 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v139/ndousse21a/ndousse21a.pdf}, url = {https://proceedings.mlr.press/v139/ndousse21a.html}, abstract = {Social learning is a key component of human and animal intelligence. By taking cues from the behavior of experts in their environment, social learners can acquire sophisticated behavior and rapidly adapt to new circumstances. This paper investigates whether independent reinforcement learning (RL) agents in a multi-agent environment can learn to use social learning to improve their performance. We find that in most circumstances, vanilla model-free RL agents do not use social learning. We analyze the reasons for this deficiency, and show that by imposing constraints on the training environment and introducing a model-based auxiliary loss we are able to obtain generalized social learning policies which enable agents to: i) discover complex skills that are not learned from single-agent training, and ii) adapt online to novel environments by taking cues from experts present in the new environment. In contrast, agents trained with model-free RL or imitation learning generalize poorly and do not succeed in the transfer tasks. By mixing multi-agent and solo training, we can obtain agents that use social learning to gain skills that they can deploy when alone, even out-performing agents trained alone from the start.} }
Endnote
%0 Conference Paper %T Emergent Social Learning via Multi-agent Reinforcement Learning %A Kamal K Ndousse %A Douglas Eck %A Sergey Levine %A Natasha Jaques %B Proceedings of the 38th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Marina Meila %E Tong Zhang %F pmlr-v139-ndousse21a %I PMLR %P 7991--8004 %U https://proceedings.mlr.press/v139/ndousse21a.html %V 139 %X Social learning is a key component of human and animal intelligence. By taking cues from the behavior of experts in their environment, social learners can acquire sophisticated behavior and rapidly adapt to new circumstances. This paper investigates whether independent reinforcement learning (RL) agents in a multi-agent environment can learn to use social learning to improve their performance. We find that in most circumstances, vanilla model-free RL agents do not use social learning. We analyze the reasons for this deficiency, and show that by imposing constraints on the training environment and introducing a model-based auxiliary loss we are able to obtain generalized social learning policies which enable agents to: i) discover complex skills that are not learned from single-agent training, and ii) adapt online to novel environments by taking cues from experts present in the new environment. In contrast, agents trained with model-free RL or imitation learning generalize poorly and do not succeed in the transfer tasks. By mixing multi-agent and solo training, we can obtain agents that use social learning to gain skills that they can deploy when alone, even out-performing agents trained alone from the start.
APA
Ndousse, K.K., Eck, D., Levine, S. & Jaques, N.. (2021). Emergent Social Learning via Multi-agent Reinforcement Learning. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:7991-8004 Available from https://proceedings.mlr.press/v139/ndousse21a.html.

Related Material