Name
Abstract | ||
|---|---|---|
We present the first massively distributed architecture for deep reinforcement learning. This architecture uses four main components: parallel actors that generate new behaviour; parallel learners that are trained from stored experience; a distributed neural network to represent the value function or behaviour policy; and a distributed store of experience. We used our architecture to implement the Deep Q-Network algorithm (DQN). Our distributed algorithm was applied to 49 games from Atari 2600 games from the Arcade Learning Environment, using identical hyperparameters. Our performance surpassed non-distributed DQN in 41 of the 49 games and also reduced the wall-time required to achieve these results by an order of magnitude on most games. |
Year | Venue | Field |
|---|---|---|
2015 | CoRR | Architecture,Hyperparameter,Massively parallel,Computer science,Bellman equation,Distributed algorithm,Artificial intelligence,Learning environment,Artificial neural network,Machine learning,Reinforcement learning |
DocType | Volume | Citations |
Journal | abs/1507.04296 | 83 |
PageRank | References | Authors |
5.85 | 9 | 14 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| arun nair | 1 | 83 | 5.85 |
| praveen srinivasan | 2 | 83 | 5.85 |
| sam blackwell | 3 | 83 | 5.85 |
| cagdas alcicek | 4 | 83 | 5.85 |
| rory fearon | 5 | 83 | 5.85 |
| alessandro de maria | 6 | 83 | 5.85 |
| vedavyas panneershelvam | 7 | 83 | 5.85 |
| mustafa suleyman | 8 | 628 | 24.43 |
| Charles Beattie | 9 | 2329 | 95.49 |
| Stig Petersen | 10 | 2329 | 95.83 |
| shane legg | 11 | 83 | 5.85 |
| volodymyr mnih | 12 | 83 | 5.85 |
| Koray Kavukcuoglu | 13 | 10189 | 504.11 |
| David Silver | 14 | 8252 | 363.86 |