Matches in SemOpenAlex for { <https://semopenalex.org/work/W4385970230> ?p ?o ?g. }
Showing items 1 to 73 of
73
with 100 items per page.
- W4385970230 abstract "Learning a policy with great generalization to unseen environments remains challenging but critical in visual reinforcement learning. Despite the success of augmentation combination in the supervised learning generalization, naively applying it to visual RL algorithms may damage the training efficiency, suffering from serve performance degradation. In this paper, we first conduct qualitative analysis and illuminate the main causes: (i) high-variance gradient magnitudes and (ii) gradient conflicts existed in various augmentation methods. To alleviate these issues, we propose a general policy gradient optimization framework, named Conflict-aware Gradient Agreement Augmentation (CG2A), and better integrate augmentation combination into visual RL algorithms to address the generalization bias. In particular, CG2A develops a Gradient Agreement Solver to adaptively balance the varying gradient magnitudes, and introduces a Soft Gradient Surgery strategy to alleviate the gradient conflicts. Extensive experiments demonstrate that CG2A significantly improves the generalization performance and sample efficiency of visual RL algorithms." @default.
- W4385970230 created "2023-08-19" @default.
- W4385970230 creator A5004300019 @default.
- W4385970230 creator A5005105786 @default.
- W4385970230 creator A5006633399 @default.
- W4385970230 creator A5011229178 @default.
- W4385970230 creator A5041083903 @default.
- W4385970230 creator A5048500768 @default.
- W4385970230 creator A5064280715 @default.
- W4385970230 creator A5064848268 @default.
- W4385970230 creator A5067539236 @default.
- W4385970230 creator A5076551144 @default.
- W4385970230 creator A5091016502 @default.
- W4385970230 date "2023-08-02" @default.
- W4385970230 modified "2023-09-27" @default.
- W4385970230 title "Improving Generalization in Visual Reinforcement Learning via Conflict-aware Gradient Agreement Augmentation" @default.
- W4385970230 doi "https://doi.org/10.48550/arxiv.2308.01194" @default.
- W4385970230 hasPublicationYear "2023" @default.
- W4385970230 type Work @default.
- W4385970230 citedByCount "0" @default.
- W4385970230 crossrefType "posted-content" @default.
- W4385970230 hasAuthorship W4385970230A5004300019 @default.
- W4385970230 hasAuthorship W4385970230A5005105786 @default.
- W4385970230 hasAuthorship W4385970230A5006633399 @default.
- W4385970230 hasAuthorship W4385970230A5011229178 @default.
- W4385970230 hasAuthorship W4385970230A5041083903 @default.
- W4385970230 hasAuthorship W4385970230A5048500768 @default.
- W4385970230 hasAuthorship W4385970230A5064280715 @default.
- W4385970230 hasAuthorship W4385970230A5064848268 @default.
- W4385970230 hasAuthorship W4385970230A5067539236 @default.
- W4385970230 hasAuthorship W4385970230A5076551144 @default.
- W4385970230 hasAuthorship W4385970230A5091016502 @default.
- W4385970230 hasBestOaLocation W43859702301 @default.
- W4385970230 hasConcept C119857082 @default.
- W4385970230 hasConcept C121955636 @default.
- W4385970230 hasConcept C134306372 @default.
- W4385970230 hasConcept C144133560 @default.
- W4385970230 hasConcept C154945302 @default.
- W4385970230 hasConcept C177148314 @default.
- W4385970230 hasConcept C196083921 @default.
- W4385970230 hasConcept C199360897 @default.
- W4385970230 hasConcept C2778770139 @default.
- W4385970230 hasConcept C33923547 @default.
- W4385970230 hasConcept C41008148 @default.
- W4385970230 hasConcept C97541855 @default.
- W4385970230 hasConceptScore W4385970230C119857082 @default.
- W4385970230 hasConceptScore W4385970230C121955636 @default.
- W4385970230 hasConceptScore W4385970230C134306372 @default.
- W4385970230 hasConceptScore W4385970230C144133560 @default.
- W4385970230 hasConceptScore W4385970230C154945302 @default.
- W4385970230 hasConceptScore W4385970230C177148314 @default.
- W4385970230 hasConceptScore W4385970230C196083921 @default.
- W4385970230 hasConceptScore W4385970230C199360897 @default.
- W4385970230 hasConceptScore W4385970230C2778770139 @default.
- W4385970230 hasConceptScore W4385970230C33923547 @default.
- W4385970230 hasConceptScore W4385970230C41008148 @default.
- W4385970230 hasConceptScore W4385970230C97541855 @default.
- W4385970230 hasLocation W43859702301 @default.
- W4385970230 hasOpenAccess W4385970230 @default.
- W4385970230 hasPrimaryLocation W43859702301 @default.
- W4385970230 hasRelatedWork W260766989 @default.
- W4385970230 hasRelatedWork W2959276766 @default.
- W4385970230 hasRelatedWork W2961085424 @default.
- W4385970230 hasRelatedWork W3074294383 @default.
- W4385970230 hasRelatedWork W3139193008 @default.
- W4385970230 hasRelatedWork W4206669594 @default.
- W4385970230 hasRelatedWork W4295941380 @default.
- W4385970230 hasRelatedWork W4306674287 @default.
- W4385970230 hasRelatedWork W4319083788 @default.
- W4385970230 hasRelatedWork W4377293004 @default.
- W4385970230 isParatext "false" @default.
- W4385970230 isRetracted "false" @default.
- W4385970230 workType "article" @default.