Statements (58)
| Predicate | Object |
|---|---|
| gptkbp:instance_of |
gptkb:software_framework
|
| gptkbp:applies_to |
gptkb:Atari_Games
|
| gptkbp:based_on |
Q-learning
|
| gptkbp:can_be_used_with |
DQN with several improvements
|
| gptkbp:developed_by |
gptkb:Google_Deep_Mind
|
| gptkbp:has_achievements |
state-of-the-art performance
|
| gptkbp:has_expansion |
gptkb:DQN
|
| gptkbp:has_variants |
Rainbow DQN with Prioritized Experience Replay
|
| https://www.w3.org/2000/01/rdf-schema#label |
Rainbow DQN
|
| gptkbp:improves |
sample efficiency
stability of training Double Q-learning Prioritized Experience Replay Distributional Reinforcement Learning Dueling Network Architecture Multi-step Learning Noisy Nets |
| gptkbp:is_associated_with |
reinforcement learning competitions
|
| gptkbp:is_cited_in |
numerous research papers
|
| gptkbp:is_compared_to |
gptkb:A3_C
TRPO PPO vanilla DQN |
| gptkbp:is_designed_for |
high-dimensional action spaces
|
| gptkbp:is_designed_to |
maximize cumulative reward
|
| gptkbp:is_evaluated_by |
mean episode returns
multiple Atari games |
| gptkbp:is_implemented_in |
gptkb:Graphics_Processing_Unit
gptkb:Py_Torch Q-learning updates |
| gptkbp:is_influenced_by |
human learning principles
|
| gptkbp:is_known_for |
its modular architecture
high performance in complex environments efficient exploration strategies robustness to hyperparameter settings |
| gptkbp:is_part_of |
the Open AI Gym
the family of DQN algorithms |
| gptkbp:is_popular_in |
benchmarking RL algorithms
|
| gptkbp:is_related_to |
Actor-Critic methods
|
| gptkbp:is_used_in |
gptkb:robot
gptkb:Research_Institute simulations game playing |
| gptkbp:is_used_to |
train agents in games
train agents in robotics train agents in simulations |
| gptkbp:performance |
other RL algorithms
|
| gptkbp:published_by |
gptkb:2017
|
| gptkbp:resulted_in |
advancements in deep learning
advancements in neural networks advancements in reinforcement learning advancements in algorithm design |
| gptkbp:suitable_for |
continuous action spaces
|
| gptkbp:training |
experience replay buffers
|
| gptkbp:uses |
gptkb:microprocessor
|
| gptkbp:utilizes |
experience replay
|
| gptkbp:bfsParent |
gptkb:Lunar_Lander-v2
|
| gptkbp:bfsLayer |
4
|