Privileged Information Dropout in Reinforcement Learning
This work addresses sample efficiency and performance in reinforcement learning, but appears incremental as it builds on existing paradigms for using privileged information.
The paper tackles the problem of leveraging privileged information during training to improve reinforcement learning agents, demonstrating that their Privileged Information Dropout method outperforms alternatives like distillation and auxiliary tasks in a simple partially-observed environment.
Using privileged information during training can improve the sample efficiency and performance of machine learning systems. This paradigm has been applied to reinforcement learning (RL), primarily in the form of distillation or auxiliary tasks, and less commonly in the form of augmenting the inputs of agents. In this work, we investigate Privileged Information Dropout (\pid) for achieving the latter which can be applied equally to value-based and policy-based RL algorithms. Within a simple partially-observed environment, we demonstrate that \pid outperforms alternatives for leveraging privileged information, including distillation and auxiliary tasks, and can successfully utilise different types of privileged information. Finally, we analyse its effect on the learned representations.