Abstract | ||
---|---|---|
Domain adaptation is an important open problem in deep reinforcement learning (RL). In many scenarios of interest data is hard to obtain, so agents may learn a source policy in a setting where data is readily available, with the hope that it generalises well to the target domain. We propose a new multi-stage RL agent, DARLA (DisentAngled Representation Learning Agent), which learns to see before learning to act. DARLAu0027s vision is based on learning a disentangled representation of the observed environment. Once DARLA can see, it is able to acquire source policies that are robust to many domain shifts - even with no access to the target domain. DARLA significantly outperforms conventional baselines in zero-shot domain adaptation scenarios, an effect that holds across a variety of RL environments (Jaco arm, DeepMind Lab) and base RL algorithms (DQN, A3C and EC). |
Year | Venue | DocType |
---|---|---|
2017 | ICML | Journal |
Volume | Citations | PageRank |
abs/1707.08475 | 19 | 0.71 |
References | Authors | |
29 | 9 |
Name | Order | Citations | PageRank |
---|---|---|---|
Irina Higgins | 1 | 245 | 11.95 |
Arka Pal | 2 | 221 | 7.85 |
Andrei A. Rusu | 3 | 19 | 0.71 |
Loïc Matthey | 4 | 239 | 10.16 |
Christopher Burgess | 5 | 236 | 9.62 |
alexander pritzel | 6 | 521 | 20.08 |
Matthew M Botvinick | 7 | 494 | 25.34 |
Charles Blundell | 8 | 822 | 41.64 |
Alexander Lerchner | 9 | 256 | 11.70 |