Learning modular structures which reflect the dynamics of the environment can lead to better generalization and robustness to changes which only affect a few of the underlying causes. We propose Recurrent Independent Mechanisms (RIMs), a new recurrent architecture in which multiple groups of recurrent cells operate with nearly independent transition dynamics, communicate only sparingly through the bottleneck of attention, and are only updated at time steps where they are most relevant. We show that this leads to specialization amongst the RIMs, which in turn allows for dramatically improved generalization on tasks where some factors of variation differ systematically between training and evaluation.
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Atari Games | Atari 2600 Beam Rider | Score | 5320 | RIMs-PPO |
| Atari Games | Atari 2600 Zaxxon | Score | 15000 | RIMs-PPO |
| Atari Games | Atari 2600 Up and Down | Score | 390000 | RIMs-PPO |
| Video Games | Atari 2600 Beam Rider | Score | 5320 | RIMs-PPO |
| Video Games | Atari 2600 Zaxxon | Score | 15000 | RIMs-PPO |
| Video Games | Atari 2600 Up and Down | Score | 390000 | RIMs-PPO |