Dueling Network Architectures

GPTKB entity

Predicate	Object
gptkbp:instanceOf	Neural network architecture
gptkbp:architecture	Deep neural network
gptkbp:arXivID	1511.06581
gptkbp:benefit	Better policy evaluation Improved learning efficiency
gptkbp:citation	5000+
gptkbp:combines	Value and advantage streams to produce Q-values
gptkbp:component	Advantage stream Value stream
gptkbp:contribution	Separates estimation of value and advantage in Q-networks
https://www.w3.org/2000/01/rdf-schema#label	Dueling Network Architectures
gptkbp:improves	Deep Q-Networks
gptkbp:introduced	gptkb:Nando_de_Freitas gptkb:Hado_van_Hasselt gptkb:Marc_Lanctot gptkb:Matteo_Hessel gptkb:Tom_Schaul gptkb:Ziyu_Wang
gptkbp:notablePublication	gptkb:Dueling_Network_Architectures_for_Deep_Reinforcement_Learning
gptkbp:openSource	gptkb:OpenAI_Baselines gptkb:Stable_Baselines gptkb:TensorFlow_Agents
gptkbp:publicationYear	2016
gptkbp:publishedIn	gptkb:ICML_2016
gptkbp:relatedTo	gptkb:Double_DQN gptkb:Prioritized_Experience_Replay
gptkbp:separates	Advantage function State-value function
gptkbp:usedFor	gptkb:Q-learning Value-based reinforcement learning
gptkbp:usedIn	Deep reinforcement learning Atari 2600 benchmarks
gptkbp:bfsParent	gptkb:Rainbow_DQN
gptkbp:bfsLayer	7