Matches in SemOpenAlex for { <https://semopenalex.org/work/W3011732969> ?p ?o ?g. }
- W3011732969 abstract "Image-based Reinforcement Learning is known to suffer from poor sample efficiency and generalisation to unseen visuals such as distractors (task-independent aspects of the observation space). Visual domain randomisation encourages transfer by training over visual factors of variation that may be encountered in the target domain. This increases learning complexity, can negatively impact learning rate and performance, and requires knowledge of potential variations during deployment. In this paper, we introduce Attention-Privileged Reinforcement Learning (APRiL) which uses a self-supervised attention mechanism to significantly alleviate these drawbacks: by focusing on task-relevant aspects of the observations, attention provides robustness to distractors as well as significantly increased learning efficiency. APRiL trains two attention-augmented actor-critic agents: one purely based on image observations, available across training and transfer domains; and one with access to privileged information (such as environment states) available only during training. Experience is shared between both agents and their attention mechanisms are aligned. The image-based policy can then be deployed without access to privileged information. We experimentally demonstrate accelerated and more robust learning on a diverse set of domains, leading to improved final performance for environments both within and outside the training distribution." @default.
- W3011732969 created "2020-03-23" @default.
- W3011732969 creator A5000964750 @default.
- W3011732969 creator A5002747297 @default.
- W3011732969 creator A5008638246 @default.
- W3011732969 creator A5027013897 @default.
- W3011732969 creator A5079415139 @default.
- W3011732969 date "2019-11-19" @default.
- W3011732969 modified "2023-10-16" @default.
- W3011732969 title "Attention-Privileged Reinforcement Learning" @default.
- W3011732969 cites W1731081199 @default.
- W3011732969 cites W2145339207 @default.
- W3011732969 cites W2155541015 @default.
- W3011732969 cites W2156737235 @default.
- W3011732969 cites W2158782408 @default.
- W3011732969 cites W2161381512 @default.
- W3011732969 cites W2165150801 @default.
- W3011732969 cites W2173248099 @default.
- W3011732969 cites W2195446438 @default.
- W3011732969 cites W2511131004 @default.
- W3011732969 cites W2529477964 @default.
- W3011732969 cites W2543474074 @default.
- W3011732969 cites W2565902248 @default.
- W3011732969 cites W2577645110 @default.
- W3011732969 cites W2605102758 @default.
- W3011732969 cites W2614839826 @default.
- W3011732969 cites W2623491082 @default.
- W3011732969 cites W2626073992 @default.
- W3011732969 cites W2733961795 @default.
- W3011732969 cites W2737215781 @default.
- W3011732969 cites W2738129230 @default.
- W3011732969 cites W2766447205 @default.
- W3011732969 cites W2767621168 @default.
- W3011732969 cites W2771807014 @default.
- W3011732969 cites W2781585732 @default.
- W3011732969 cites W2900152462 @default.
- W3011732969 cites W2914688076 @default.
- W3011732969 cites W2949924544 @default.
- W3011732969 cites W2951670162 @default.
- W3011732969 cites W2952629144 @default.
- W3011732969 cites W2963175324 @default.
- W3011732969 cites W2963390419 @default.
- W3011732969 cites W2964198579 @default.
- W3011732969 cites W2970909667 @default.
- W3011732969 cites W2981030070 @default.
- W3011732969 cites W2982187448 @default.
- W3011732969 cites W2990747716 @default.
- W3011732969 cites W3009492911 @default.
- W3011732969 cites W3021708257 @default.
- W3011732969 cites W3101442004 @default.
- W3011732969 cites W3108330043 @default.
- W3011732969 cites W3115293622 @default.
- W3011732969 hasPublicationYear "2019" @default.
- W3011732969 type Work @default.
- W3011732969 sameAs 3011732969 @default.
- W3011732969 citedByCount "3" @default.
- W3011732969 countsByYear W30117329692020 @default.
- W3011732969 countsByYear W30117329692021 @default.
- W3011732969 crossrefType "posted-content" @default.
- W3011732969 hasAuthorship W3011732969A5000964750 @default.
- W3011732969 hasAuthorship W3011732969A5002747297 @default.
- W3011732969 hasAuthorship W3011732969A5008638246 @default.
- W3011732969 hasAuthorship W3011732969A5027013897 @default.
- W3011732969 hasAuthorship W3011732969A5079415139 @default.
- W3011732969 hasConcept C104317684 @default.
- W3011732969 hasConcept C105339364 @default.
- W3011732969 hasConcept C107457646 @default.
- W3011732969 hasConcept C111919701 @default.
- W3011732969 hasConcept C119857082 @default.
- W3011732969 hasConcept C138885662 @default.
- W3011732969 hasConcept C150899416 @default.
- W3011732969 hasConcept C154945302 @default.
- W3011732969 hasConcept C162324750 @default.
- W3011732969 hasConcept C171041071 @default.
- W3011732969 hasConcept C177264268 @default.
- W3011732969 hasConcept C185592680 @default.
- W3011732969 hasConcept C187736073 @default.
- W3011732969 hasConcept C199360897 @default.
- W3011732969 hasConcept C2779178101 @default.
- W3011732969 hasConcept C2780451532 @default.
- W3011732969 hasConcept C41008148 @default.
- W3011732969 hasConcept C41895202 @default.
- W3011732969 hasConcept C55493867 @default.
- W3011732969 hasConcept C63479239 @default.
- W3011732969 hasConcept C97541855 @default.
- W3011732969 hasConceptScore W3011732969C104317684 @default.
- W3011732969 hasConceptScore W3011732969C105339364 @default.
- W3011732969 hasConceptScore W3011732969C107457646 @default.
- W3011732969 hasConceptScore W3011732969C111919701 @default.
- W3011732969 hasConceptScore W3011732969C119857082 @default.
- W3011732969 hasConceptScore W3011732969C138885662 @default.
- W3011732969 hasConceptScore W3011732969C150899416 @default.
- W3011732969 hasConceptScore W3011732969C154945302 @default.
- W3011732969 hasConceptScore W3011732969C162324750 @default.
- W3011732969 hasConceptScore W3011732969C171041071 @default.
- W3011732969 hasConceptScore W3011732969C177264268 @default.
- W3011732969 hasConceptScore W3011732969C185592680 @default.
- W3011732969 hasConceptScore W3011732969C187736073 @default.
- W3011732969 hasConceptScore W3011732969C199360897 @default.
- W3011732969 hasConceptScore W3011732969C2779178101 @default.