Corpus ID: 231924682

Domain Adversarial Reinforcement Learning

@article{Li2021DomainAR,
  title={Domain Adversarial Reinforcement Learning},
  author={Bonnie Li and Vincent Franccois-Lavet and T. Doan and Joelle Pineau},
  journal={ArXiv},
  year={2021},
  volume={abs/2102.07097}
}
We consider the problem of generalization in reinforcement learning where visual aspects of the observations might differ, e.g. when there are different backgrounds or change in contrast, brightness, etc. We assume that our agent has access to only a few of the MDPs from the MDP distribution during training. The performance of the agent is then reported on new unknown test domains drawn from the distribution (e.g. unseen backgrounds). For this “zero-shot RL” task, we enforce invariance of the… Expand

References

SHOWING 1-10 OF 32 REFERENCES
Visual Transfer for Reinforcement Learning via Wasserstein Domain Confusion
  • 5
  • Highly Influential
  • PDF
Adversarial Discriminative Domain Adaptation
  • 1,960
  • PDF
Quantifying Generalization in Reinforcement Learning
  • 216
  • PDF
Deep Transfer Learning with Joint Adaptation Networks
  • 860
  • PDF
Self-Supervised Policy Adaptation during Deployment
  • 10
  • Highly Influential
  • PDF
Deep Reinforcement and InfoMax Learning
  • 9
  • PDF
Natural Environment Benchmarks for Reinforcement Learning
  • 30
  • PDF
Actor-Mimic: Deep Multitask and Transfer Reinforcement Learning
  • 350
  • PDF
Unsupervised Domain Adaptation by Backpropagation
  • 2,156
  • Highly Influential
  • PDF
...
1
2
3
4
...