Prioritized Experience Replay

GPTKB entity

Statements (19)
Predicate Object
gptkbp:instanceOf reinforcement learning technique
gptkbp:address inefficiency of uniform sampling
gptkbp:amendedBy experience replay
gptkbp:appliesTo gptkb:DQN
gptkb:Rainbow_DQN
gptkbp:category gptkb:model
gptkb:reinforcement_learning
gptkbp:citation over 4000
gptkbp:hasConcept sampling probability proportional to TD error
https://www.w3.org/2000/01/rdf-schema#label Prioritized Experience Replay
gptkbp:improves sample efficiency
learning speed
gptkbp:priorityDate transitions with high TD error
gptkbp:proposedBy gptkb:Tom_Schaul
2015
gptkbp:publishedIn arXiv:1511.05952
gptkbp:usedIn deep reinforcement learning
gptkbp:bfsParent gptkb:Deep_Q-Network_(DQN)
gptkbp:bfsLayer 6