Matches in SemOpenAlex for { <https://semopenalex.org/work/W4304098570> ?p ?o ?g. }
Showing items 1 to 82 of
82
with 100 items per page.
- W4304098570 abstract "In recent years, deep learning has achieved a great success in various image recognition tasks. However, the long-tailed setting over a semantic class plays a leading role in real-world applications. Common methods focus on optimization on balanced distribution or naive models. Few works explore long-tailed learning from a deep learning-based generalization perspective. The loss landscape on long-tailed learning is first investigated in this work. Empirical results show that sharpness-aware optimizers work not well on long-tailed learning. Because they do not take class priors into consideration, and they fail to improve performance of few-shot classes. To better guide the network and explicitly alleviate sharpness without extra computational burden, we develop a universal Balanced Gradient Penalty (BGP) method. Surprisingly, our BGP method does not need the detailed class priors and preserves privacy. Our new algorithm BGP, as a regularization loss, can achieve the state-of-the-art results on various image datasets (i.e., CIFAR-LT, ImageNet-LT and iNaturalist-2018) in the settings of different imbalance ratios." @default.
- W4304098570 created "2022-10-10" @default.
- W4304098570 creator A5045129199 @default.
- W4304098570 creator A5046454314 @default.
- W4304098570 creator A5054464310 @default.
- W4304098570 creator A5065691762 @default.
- W4304098570 creator A5067858777 @default.
- W4304098570 creator A5077622519 @default.
- W4304098570 date "2022-10-10" @default.
- W4304098570 modified "2023-10-16" @default.
- W4304098570 title "Balanced Gradient Penalty Improves Deep Long-Tailed Learning" @default.
- W4304098570 cites W1536680647 @default.
- W4304098570 cites W2029029543 @default.
- W4304098570 cites W2108598243 @default.
- W4304098570 cites W2148143831 @default.
- W4304098570 cites W2194775991 @default.
- W4304098570 cites W2552194003 @default.
- W4304098570 cites W2794704013 @default.
- W4304098570 cites W2962933664 @default.
- W4304098570 cites W2963351448 @default.
- W4304098570 cites W2963691377 @default.
- W4304098570 cites W3034601242 @default.
- W4304098570 cites W3034711780 @default.
- W4304098570 cites W3035552357 @default.
- W4304098570 cites W3096121526 @default.
- W4304098570 cites W3166792421 @default.
- W4304098570 cites W3175613352 @default.
- W4304098570 cites W3202232857 @default.
- W4304098570 doi "https://doi.org/10.1145/3503161.3547763" @default.
- W4304098570 hasPublicationYear "2022" @default.
- W4304098570 type Work @default.
- W4304098570 citedByCount "1" @default.
- W4304098570 countsByYear W43040985702023 @default.
- W4304098570 crossrefType "proceedings-article" @default.
- W4304098570 hasAuthorship W4304098570A5045129199 @default.
- W4304098570 hasAuthorship W4304098570A5046454314 @default.
- W4304098570 hasAuthorship W4304098570A5054464310 @default.
- W4304098570 hasAuthorship W4304098570A5065691762 @default.
- W4304098570 hasAuthorship W4304098570A5067858777 @default.
- W4304098570 hasAuthorship W4304098570A5077622519 @default.
- W4304098570 hasConcept C107673813 @default.
- W4304098570 hasConcept C108583219 @default.
- W4304098570 hasConcept C115961682 @default.
- W4304098570 hasConcept C119857082 @default.
- W4304098570 hasConcept C120665830 @default.
- W4304098570 hasConcept C121332964 @default.
- W4304098570 hasConcept C12713177 @default.
- W4304098570 hasConcept C154945302 @default.
- W4304098570 hasConcept C177769412 @default.
- W4304098570 hasConcept C192209626 @default.
- W4304098570 hasConcept C2776135515 @default.
- W4304098570 hasConcept C2777212361 @default.
- W4304098570 hasConcept C41008148 @default.
- W4304098570 hasConceptScore W4304098570C107673813 @default.
- W4304098570 hasConceptScore W4304098570C108583219 @default.
- W4304098570 hasConceptScore W4304098570C115961682 @default.
- W4304098570 hasConceptScore W4304098570C119857082 @default.
- W4304098570 hasConceptScore W4304098570C120665830 @default.
- W4304098570 hasConceptScore W4304098570C121332964 @default.
- W4304098570 hasConceptScore W4304098570C12713177 @default.
- W4304098570 hasConceptScore W4304098570C154945302 @default.
- W4304098570 hasConceptScore W4304098570C177769412 @default.
- W4304098570 hasConceptScore W4304098570C192209626 @default.
- W4304098570 hasConceptScore W4304098570C2776135515 @default.
- W4304098570 hasConceptScore W4304098570C2777212361 @default.
- W4304098570 hasConceptScore W4304098570C41008148 @default.
- W4304098570 hasLocation W43040985701 @default.
- W4304098570 hasOpenAccess W4304098570 @default.
- W4304098570 hasPrimaryLocation W43040985701 @default.
- W4304098570 hasRelatedWork W3014300295 @default.
- W4304098570 hasRelatedWork W3164822677 @default.
- W4304098570 hasRelatedWork W4223943233 @default.
- W4304098570 hasRelatedWork W4225161397 @default.
- W4304098570 hasRelatedWork W4250304930 @default.
- W4304098570 hasRelatedWork W4312200629 @default.
- W4304098570 hasRelatedWork W4360585206 @default.
- W4304098570 hasRelatedWork W4364306694 @default.
- W4304098570 hasRelatedWork W4380075502 @default.
- W4304098570 hasRelatedWork W4380086463 @default.
- W4304098570 isParatext "false" @default.
- W4304098570 isRetracted "false" @default.
- W4304098570 workType "article" @default.