Matches in SemOpenAlex for { <https://semopenalex.org/work/W3118683362> ?p ?o ?g. }
Showing items 1 to 78 of
78
with 100 items per page.
- W3118683362 abstract "Top-performing approaches to embodied AI tasks like point-goal navigation often rely on training agents via reinforcement learning over tens of millions (or even billions) of experiential steps - learning neural agents that map directly from visual observations to actions. In this work, we question whether these extreme training durations are necessary or if they are simply due to the difficulty of learning visual representations purely from task reward. We examine the task of point-goal navigation in photorealistic environments and introduce three auxiliary tasks that encourage learned representations to capture key elements of the task - local scene geometry, transition dynamics of the environment, and progress towards the goal. Importantly, these can be evaluated independent of task performance and provide strong supervision for representation learning. Our auxiliary tasks are simple to implement and rely on supervision already present in simulators commonly used for point-goal navigation. Applying our auxiliary losses to agents from prior works, we observe a >4 × improvement in sample efficiency - in 17 million steps, our augmented agents outperforms state-of-the-art agents trained for 72 million steps." @default.
- W3118683362 created "2021-01-18" @default.
- W3118683362 creator A5024698329 @default.
- W3118683362 creator A5051259505 @default.
- W3118683362 date "2021-01-01" @default.
- W3118683362 modified "2023-10-06" @default.
- W3118683362 title "Auxiliary Tasks for Efficient Learning of Point-Goal Navigation" @default.
- W3118683362 cites W2157331557 @default.
- W3118683362 cites W2198618282 @default.
- W3118683362 cites W2549139847 @default.
- W3118683362 cites W2593841437 @default.
- W3118683362 cites W2930283066 @default.
- W3118683362 cites W2953127211 @default.
- W3118683362 cites W2963800628 @default.
- W3118683362 cites W2964339842 @default.
- W3118683362 cites W3009928773 @default.
- W3118683362 cites W3109085430 @default.
- W3118683362 doi "https://doi.org/10.1109/wacv48630.2021.00076" @default.
- W3118683362 hasPublicationYear "2021" @default.
- W3118683362 type Work @default.
- W3118683362 sameAs 3118683362 @default.
- W3118683362 citedByCount "2" @default.
- W3118683362 countsByYear W31186833622021 @default.
- W3118683362 countsByYear W31186833622023 @default.
- W3118683362 crossrefType "proceedings-article" @default.
- W3118683362 hasAuthorship W3118683362A5024698329 @default.
- W3118683362 hasAuthorship W3118683362A5051259505 @default.
- W3118683362 hasConcept C100609095 @default.
- W3118683362 hasConcept C107457646 @default.
- W3118683362 hasConcept C119857082 @default.
- W3118683362 hasConcept C154945302 @default.
- W3118683362 hasConcept C162324750 @default.
- W3118683362 hasConcept C17744445 @default.
- W3118683362 hasConcept C187736073 @default.
- W3118683362 hasConcept C199539241 @default.
- W3118683362 hasConcept C2524010 @default.
- W3118683362 hasConcept C2776359362 @default.
- W3118683362 hasConcept C2780451532 @default.
- W3118683362 hasConcept C28719098 @default.
- W3118683362 hasConcept C33923547 @default.
- W3118683362 hasConcept C37228920 @default.
- W3118683362 hasConcept C41008148 @default.
- W3118683362 hasConcept C94625758 @default.
- W3118683362 hasConcept C97541855 @default.
- W3118683362 hasConceptScore W3118683362C100609095 @default.
- W3118683362 hasConceptScore W3118683362C107457646 @default.
- W3118683362 hasConceptScore W3118683362C119857082 @default.
- W3118683362 hasConceptScore W3118683362C154945302 @default.
- W3118683362 hasConceptScore W3118683362C162324750 @default.
- W3118683362 hasConceptScore W3118683362C17744445 @default.
- W3118683362 hasConceptScore W3118683362C187736073 @default.
- W3118683362 hasConceptScore W3118683362C199539241 @default.
- W3118683362 hasConceptScore W3118683362C2524010 @default.
- W3118683362 hasConceptScore W3118683362C2776359362 @default.
- W3118683362 hasConceptScore W3118683362C2780451532 @default.
- W3118683362 hasConceptScore W3118683362C28719098 @default.
- W3118683362 hasConceptScore W3118683362C33923547 @default.
- W3118683362 hasConceptScore W3118683362C37228920 @default.
- W3118683362 hasConceptScore W3118683362C41008148 @default.
- W3118683362 hasConceptScore W3118683362C94625758 @default.
- W3118683362 hasConceptScore W3118683362C97541855 @default.
- W3118683362 hasLocation W31186833621 @default.
- W3118683362 hasOpenAccess W3118683362 @default.
- W3118683362 hasPrimaryLocation W31186833621 @default.
- W3118683362 hasRelatedWork W2031296774 @default.
- W3118683362 hasRelatedWork W2325123304 @default.
- W3118683362 hasRelatedWork W2959276766 @default.
- W3118683362 hasRelatedWork W2961085424 @default.
- W3118683362 hasRelatedWork W3074294383 @default.
- W3118683362 hasRelatedWork W4206669594 @default.
- W3118683362 hasRelatedWork W4295941380 @default.
- W3118683362 hasRelatedWork W4312391057 @default.
- W3118683362 hasRelatedWork W4319083788 @default.
- W3118683362 hasRelatedWork W4327926809 @default.
- W3118683362 isParatext "false" @default.
- W3118683362 isRetracted "false" @default.
- W3118683362 magId "3118683362" @default.
- W3118683362 workType "article" @default.