Matches in SemOpenAlex for { <https://semopenalex.org/work/W2893662673> ?p ?o ?g. }
Showing items 1 to 94 of
94
with 100 items per page.
- W2893662673 abstract "Deep reinforcement learning algorithms have shown an impressive ability to learn complex control policies in high-dimensional tasks. However, despite the ever-increasing performance on popular benchmarks, policies learned by deep reinforcement learning algorithms can struggle to generalize when evaluated in remarkably similar environments. In this paper we propose a protocol to evaluate generalization in reinforcement learning through different modes of Atari 2600 games. With that protocol we assess the generalization capabilities of DQN, one of the most traditional deep reinforcement learning algorithms, and we provide evidence suggesting that DQN overspecializes to the training environment. We then comprehensively evaluate the impact of dropout and $ell_2$ regularization, as well as the impact of reusing learned representations to improve the generalization capabilities of DQN. Despite regularization being largely underutilized in deep reinforcement learning, we show that it can, in fact, help DQN learn more general features. These features can be reused and fine-tuned on similar tasks, considerably improving DQN's sample efficiency." @default.
- W2893662673 created "2018-10-05" @default.
- W2893662673 creator A5047755961 @default.
- W2893662673 creator A5081163135 @default.
- W2893662673 creator A5085413987 @default.
- W2893662673 date "2018-09-27" @default.
- W2893662673 modified "2023-09-27" @default.
- W2893662673 title "Generalization and Regularization in DQN" @default.
- W2893662673 cites W1533861849 @default.
- W2893662673 cites W1903029394 @default.
- W2893662673 cites W2095705004 @default.
- W2893662673 cites W2104228245 @default.
- W2893662673 cites W2124175081 @default.
- W2893662673 cites W2145339207 @default.
- W2893662673 cites W2149933564 @default.
- W2893662673 cites W2158738729 @default.
- W2893662673 cites W2174786457 @default.
- W2893662673 cites W2504108613 @default.
- W2893662673 cites W2560647685 @default.
- W2893662673 cites W2754517384 @default.
- W2893662673 cites W2786036274 @default.
- W2893662673 cites W2796979132 @default.
- W2893662673 cites W2903181768 @default.
- W2893662673 cites W2904815624 @default.
- W2893662673 cites W2914898814 @default.
- W2893662673 cites W2950928354 @default.
- W2893662673 cites W2962749646 @default.
- W2893662673 cites W2963199420 @default.
- W2893662673 cites W2963403143 @default.
- W2893662673 cites W2964001908 @default.
- W2893662673 cites W2964352358 @default.
- W2893662673 cites W2426267443 @default.
- W2893662673 hasPublicationYear "2018" @default.
- W2893662673 type Work @default.
- W2893662673 sameAs 2893662673 @default.
- W2893662673 citedByCount "65" @default.
- W2893662673 countsByYear W28936626732018 @default.
- W2893662673 countsByYear W28936626732019 @default.
- W2893662673 countsByYear W28936626732020 @default.
- W2893662673 countsByYear W28936626732021 @default.
- W2893662673 countsByYear W28936626732022 @default.
- W2893662673 crossrefType "posted-content" @default.
- W2893662673 hasAuthorship W2893662673A5047755961 @default.
- W2893662673 hasAuthorship W2893662673A5081163135 @default.
- W2893662673 hasAuthorship W2893662673A5085413987 @default.
- W2893662673 hasConcept C119857082 @default.
- W2893662673 hasConcept C134306372 @default.
- W2893662673 hasConcept C154945302 @default.
- W2893662673 hasConcept C15744967 @default.
- W2893662673 hasConcept C177148314 @default.
- W2893662673 hasConcept C2776135515 @default.
- W2893662673 hasConcept C33923547 @default.
- W2893662673 hasConcept C41008148 @default.
- W2893662673 hasConcept C67203356 @default.
- W2893662673 hasConcept C77805123 @default.
- W2893662673 hasConcept C97541855 @default.
- W2893662673 hasConceptScore W2893662673C119857082 @default.
- W2893662673 hasConceptScore W2893662673C134306372 @default.
- W2893662673 hasConceptScore W2893662673C154945302 @default.
- W2893662673 hasConceptScore W2893662673C15744967 @default.
- W2893662673 hasConceptScore W2893662673C177148314 @default.
- W2893662673 hasConceptScore W2893662673C2776135515 @default.
- W2893662673 hasConceptScore W2893662673C33923547 @default.
- W2893662673 hasConceptScore W2893662673C41008148 @default.
- W2893662673 hasConceptScore W2893662673C67203356 @default.
- W2893662673 hasConceptScore W2893662673C77805123 @default.
- W2893662673 hasConceptScore W2893662673C97541855 @default.
- W2893662673 hasLocation W28936626731 @default.
- W2893662673 hasOpenAccess W2893662673 @default.
- W2893662673 hasPrimaryLocation W28936626731 @default.
- W2893662673 hasRelatedWork W1757796397 @default.
- W2893662673 hasRelatedWork W1771410628 @default.
- W2893662673 hasRelatedWork W2121863487 @default.
- W2893662673 hasRelatedWork W2145339207 @default.
- W2893662673 hasRelatedWork W2194775991 @default.
- W2893662673 hasRelatedWork W2605102758 @default.
- W2893662673 hasRelatedWork W2736601468 @default.
- W2893662673 hasRelatedWork W2786036274 @default.
- W2893662673 hasRelatedWork W2797527950 @default.
- W2893662673 hasRelatedWork W2809668646 @default.
- W2893662673 hasRelatedWork W2891790128 @default.
- W2893662673 hasRelatedWork W2898436992 @default.
- W2893662673 hasRelatedWork W2963403143 @default.
- W2893662673 hasRelatedWork W2963680188 @default.
- W2893662673 hasRelatedWork W2964043796 @default.
- W2893662673 hasRelatedWork W2964121744 @default.
- W2893662673 hasRelatedWork W2970214542 @default.
- W2893662673 hasRelatedWork W2999617596 @default.
- W2893662673 hasRelatedWork W3021708257 @default.
- W2893662673 hasRelatedWork W3103780890 @default.
- W2893662673 isParatext "false" @default.
- W2893662673 isRetracted "false" @default.
- W2893662673 magId "2893662673" @default.
- W2893662673 workType "article" @default.