Real-Time Recurrent Learning using Trace Units in Reinforcement Learning
This work addresses the computational bottleneck of real-time recurrent learning for agents in partially observable environments, offering an incremental improvement over linear recurrent architectures.
The paper tackles the problem of training recurrent neural networks efficiently in online reinforcement learning for partially observable environments by introducing Recurrent Trace Units (RTUs), which outperform other architectures with less computation.
Recurrent Neural Networks (RNNs) are used to learn representations in partially observable environments. For agents that learn online and continually interact with the environment, it is desirable to train RNNs with real-time recurrent learning (RTRL); unfortunately, RTRL is prohibitively expensive for standard RNNs. A promising direction is to use linear recurrent architectures (LRUs), where dense recurrent weights are replaced with a complex-valued diagonal, making RTRL efficient. In this work, we build on these insights to provide a lightweight but effective approach for training RNNs in online RL. We introduce Recurrent Trace Units (RTUs), a small modification on LRUs that we nonetheless find to have significant performance benefits over LRUs when trained with RTRL. We find RTUs significantly outperform other recurrent architectures across several partially observable environments while using significantly less computation.