Matches in SemOpenAlex for { <https://semopenalex.org/work/W4200301265> ?p ?o ?g. }
Showing items 1 to 72 of
72
with 100 items per page.
- W4200301265 abstract "<p>Long-tailed data is still a big challenge for deep neural networks, even though they have achieved great success on balanced data. We observe that vanilla training on long-tailed data with cross-entropy loss makes the instance-rich head classes severely squeeze the spatial distribution of the tail classes, which leads to difficulty in classifying tail class samples. Furthermore, the original cross-entropy loss can only propagate gradient short-lively because the gradient in softmax form rapidly approaches zero as the logit difference increases. This phenomenon is called softmax saturation. It is unfavorable for training on balanced data, but can be utilized to adjust the validity of the samples in long-tailed data, thereby solving the distorted embedding space of long-tailed problems. To this end, this paper therefore proposes the Gaussian clouded logit adjustment by Gaussian perturbing different class logits with varied amplitude. We define the amplitude of perturbation as cloud size and set relatively large cloud sizes to tail classes. The large cloud size can reduce the softmax saturation and thereby making tail class samples more active as well as enlarging the embedding space. To alleviate the bias in the classifier, we accordingly propose the class-based effective number sampling strategy with classifier re-training. Extensive experiments on benchmark datasets validate the superior performance of the proposed method.</p><br>" @default.
- W4200301265 created "2021-12-31" @default.
- W4200301265 creator A5038516431 @default.
- W4200301265 creator A5044521860 @default.
- W4200301265 creator A5064228800 @default.
- W4200301265 date "2021-11-19" @default.
- W4200301265 modified "2023-10-16" @default.
- W4200301265 title "Long-tailed Visual Recognition via Gaussian Clouded Logit Adjustment" @default.
- W4200301265 doi "https://doi.org/10.36227/techrxiv.17031920.v1" @default.
- W4200301265 hasPublicationYear "2021" @default.
- W4200301265 type Work @default.
- W4200301265 citedByCount "0" @default.
- W4200301265 crossrefType "posted-content" @default.
- W4200301265 hasAuthorship W4200301265A5038516431 @default.
- W4200301265 hasAuthorship W4200301265A5044521860 @default.
- W4200301265 hasAuthorship W4200301265A5064228800 @default.
- W4200301265 hasBestOaLocation W42003012651 @default.
- W4200301265 hasConcept C105795698 @default.
- W4200301265 hasConcept C106301342 @default.
- W4200301265 hasConcept C11413529 @default.
- W4200301265 hasConcept C119857082 @default.
- W4200301265 hasConcept C121332964 @default.
- W4200301265 hasConcept C140331021 @default.
- W4200301265 hasConcept C153180895 @default.
- W4200301265 hasConcept C154945302 @default.
- W4200301265 hasConcept C163716315 @default.
- W4200301265 hasConcept C180205008 @default.
- W4200301265 hasConcept C188441871 @default.
- W4200301265 hasConcept C197640229 @default.
- W4200301265 hasConcept C33923547 @default.
- W4200301265 hasConcept C41008148 @default.
- W4200301265 hasConcept C41608201 @default.
- W4200301265 hasConcept C50644808 @default.
- W4200301265 hasConcept C51632099 @default.
- W4200301265 hasConcept C62520636 @default.
- W4200301265 hasConcept C95623464 @default.
- W4200301265 hasConceptScore W4200301265C105795698 @default.
- W4200301265 hasConceptScore W4200301265C106301342 @default.
- W4200301265 hasConceptScore W4200301265C11413529 @default.
- W4200301265 hasConceptScore W4200301265C119857082 @default.
- W4200301265 hasConceptScore W4200301265C121332964 @default.
- W4200301265 hasConceptScore W4200301265C140331021 @default.
- W4200301265 hasConceptScore W4200301265C153180895 @default.
- W4200301265 hasConceptScore W4200301265C154945302 @default.
- W4200301265 hasConceptScore W4200301265C163716315 @default.
- W4200301265 hasConceptScore W4200301265C180205008 @default.
- W4200301265 hasConceptScore W4200301265C188441871 @default.
- W4200301265 hasConceptScore W4200301265C197640229 @default.
- W4200301265 hasConceptScore W4200301265C33923547 @default.
- W4200301265 hasConceptScore W4200301265C41008148 @default.
- W4200301265 hasConceptScore W4200301265C41608201 @default.
- W4200301265 hasConceptScore W4200301265C50644808 @default.
- W4200301265 hasConceptScore W4200301265C51632099 @default.
- W4200301265 hasConceptScore W4200301265C62520636 @default.
- W4200301265 hasConceptScore W4200301265C95623464 @default.
- W4200301265 hasLocation W42003012651 @default.
- W4200301265 hasLocation W42003012652 @default.
- W4200301265 hasOpenAccess W4200301265 @default.
- W4200301265 hasPrimaryLocation W42003012651 @default.
- W4200301265 hasRelatedWork W10719664 @default.
- W4200301265 hasRelatedWork W10917997 @default.
- W4200301265 hasRelatedWork W10934940 @default.
- W4200301265 hasRelatedWork W11506839 @default.
- W4200301265 hasRelatedWork W3007262 @default.
- W4200301265 hasRelatedWork W3030276 @default.
- W4200301265 hasRelatedWork W3968525 @default.
- W4200301265 hasRelatedWork W5979161 @default.
- W4200301265 hasRelatedWork W7591744 @default.
- W4200301265 hasRelatedWork W7655147 @default.
- W4200301265 isParatext "false" @default.
- W4200301265 isRetracted "false" @default.
- W4200301265 workType "article" @default.