Improving Generalization in Reinforcement Learning with Mixture Regularization
This work addresses generalization issues in reinforcement learning for agents deployed in diverse environments, representing an incremental improvement over existing data augmentation techniques.
The paper tackles the problem of deep reinforcement learning agents overfitting to limited training environments and failing to generalize to unseen ones, by introducing mixreg, a method that trains agents on mixtures of observations from different environments with linearity constraints, resulting in large-margin performance improvements on the Procgen benchmark.
Deep reinforcement learning (RL) agents trained in a limited set of environments tend to suffer overfitting and fail to generalize to unseen testing environments. To improve their generalizability, data augmentation approaches (e.g. cutout and random convolution) are previously explored to increase the data diversity. However, we find these approaches only locally perturb the observations regardless of the training environments, showing limited effectiveness on enhancing the data diversity and the generalization performance. In this work, we introduce a simple approach, named mixreg, which trains agents on a mixture of observations from different training environments and imposes linearity constraints on the observation interpolations and the supervision (e.g. associated reward) interpolations. Mixreg increases the data diversity more effectively and helps learn smoother policies. We verify its effectiveness on improving generalization by conducting extensive experiments on the large-scale Procgen benchmark. Results show mixreg outperforms the well-established baselines on unseen testing environments by a large margin. Mixreg is simple, effective and general. It can be applied to both policy-based and value-based RL algorithms. Code is available at https://github.com/kaixin96/mixreg .