Improving Generalization in Meta-RL with Imaginary Tasks from Latent Dynamics Mixture

By Sae-Young Chung. 2021

22 citations recorded by the source corpus.

The generalization ability of most meta-reinforcement learning (meta-RL) methods is largely limited to test tasks that are sampled from the same distribution used to sample training tasks. To overcome the limitation, we propose Latent Dynamics Mixture (LDM) that trains a reinforcement learning agent with imaginary tasks generated from mixtures of learned latent dynamics. By training a policy on mixture tasks along with original training tasks, LDM allows the agent to prepare for unseen test tasks during training and prevents the agent from overfitting the training tasks. LDM significantly outperforms standard meta-RL methods in test returns on the gridworld navigation and MuJoCo tasks where we strictly separate the training task distribution and the test task distribution.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Matthew Riemer
2. Irina Rish
3. Khimya Khetarpal
4. Doina Precup
5. Sergey Levine
6. Joelle Pineau
7. Eric Eaton
8. Chelsea Finn

In Computer Vision and Pattern Recognition

1. Matthew Riemer
2. Irina Rish
3. Doina Precup
4. Khimya Khetarpal
5. Lucas Caccia

Citations in the corpus

• Towards Continual Reinforcement Learning: A Review and Perspectives (2020, background): “For example, agents have in some settings learned to generate their own imagined environments or scenarios for future play (Wang, Lehman, Clune, and Stanley, 2019; Raileanu, Goldstein, Yarats, Kostrikov, and Fergus, 2021; Chen, Zhang, Xu, Ma, Yang, Song, Wang, and Wu, 2021; Lee and Chung, 2021), the”