Improving Generalization in Meta-RL with Imaginary Tasks from Latent Dynamics Mixture
This addresses the challenge of overfitting in meta-RL for researchers and practitioners, offering a method to enhance generalization to unseen tasks, though it appears incremental as it builds on existing meta-RL frameworks.
The paper tackles the problem of limited generalization in meta-reinforcement learning by proposing Latent Dynamics Mixture (LDM), which generates imaginary tasks from mixtures of learned latent dynamics to train agents, resulting in significantly improved test returns on gridworld navigation and MuJoCo tasks with strict separation between training and test distributions.
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.