Matches in SemOpenAlex for { <https://semopenalex.org/work/W3174917907> ?p ?o ?g. }
Showing items 1 to 86 of
86
with 100 items per page.
- W3174917907 abstract "Online class-incremental continual learning (CL) studies the problem of learning new classes continually from an online non-stationary data stream, intending to adapt to new data while mitigating catastrophic forgetting. While memory replay has shown promising results, the recency bias in online learning caused by the commonly used Softmax classifier remains an unsolved challenge. Although the Nearest-Class-Mean (NCM) classifier is significantly undervalued in the CL community, we demonstrate that it is a simple yet effective substitute for the Softmax classifier. It addresses the recency bias and avoids structural changes in the fully-connected layer for new classes. Moreover, we observe considerable and consistent performance gains when replacing the Softmax classifier with the NCM classifier for several state-of-the-art replay methods.To leverage the NCM classifier more effectively, data embeddings belonging to the same class should be clustered and well-separated from those with a different class label. To this end, we contribute Supervised Contrastive Replay (SCR), which explicitly encourages samples from the same class to cluster tightly in embedding space while pushing those of different classes further apart during replay-based training. Overall, we observe that our proposed SCR substantially reduces catastrophic forgetting and outperforms state-of-the-art CL methods by a significant margin on a variety of datasets." @default.
- W3174917907 created "2021-07-05" @default.
- W3174917907 creator A5008383067 @default.
- W3174917907 creator A5028174137 @default.
- W3174917907 creator A5062110178 @default.
- W3174917907 creator A5066572795 @default.
- W3174917907 date "2021-06-01" @default.
- W3174917907 modified "2023-10-15" @default.
- W3174917907 title "Supervised Contrastive Replay: Revisiting the Nearest Class Mean Classifier in Online Class-Incremental Continual Learning" @default.
- W3174917907 cites W1682403713 @default.
- W3174917907 cites W2015563892 @default.
- W3174917907 cites W2119885577 @default.
- W3174917907 cites W2560647685 @default.
- W3174917907 cites W2788388592 @default.
- W3174917907 cites W2798836702 @default.
- W3174917907 cites W2798991696 @default.
- W3174917907 cites W2948734064 @default.
- W3174917907 cites W2954929116 @default.
- W3174917907 cites W2963072899 @default.
- W3174917907 cites W2963115079 @default.
- W3174917907 cites W2963499153 @default.
- W3174917907 cites W2973049979 @default.
- W3174917907 cites W2981864462 @default.
- W3174917907 cites W3034942609 @default.
- W3174917907 cites W3035501943 @default.
- W3174917907 cites W3035524453 @default.
- W3174917907 cites W3096655658 @default.
- W3174917907 cites W3099152386 @default.
- W3174917907 cites W3103800629 @default.
- W3174917907 cites W3114632476 @default.
- W3174917907 doi "https://doi.org/10.1109/cvprw53098.2021.00398" @default.
- W3174917907 hasPublicationYear "2021" @default.
- W3174917907 type Work @default.
- W3174917907 sameAs 3174917907 @default.
- W3174917907 citedByCount "38" @default.
- W3174917907 countsByYear W31749179072021 @default.
- W3174917907 countsByYear W31749179072022 @default.
- W3174917907 countsByYear W31749179072023 @default.
- W3174917907 crossrefType "proceedings-article" @default.
- W3174917907 hasAuthorship W3174917907A5008383067 @default.
- W3174917907 hasAuthorship W3174917907A5028174137 @default.
- W3174917907 hasAuthorship W3174917907A5062110178 @default.
- W3174917907 hasAuthorship W3174917907A5066572795 @default.
- W3174917907 hasBestOaLocation W31749179072 @default.
- W3174917907 hasConcept C113238511 @default.
- W3174917907 hasConcept C119857082 @default.
- W3174917907 hasConcept C138885662 @default.
- W3174917907 hasConcept C153083717 @default.
- W3174917907 hasConcept C153180895 @default.
- W3174917907 hasConcept C154945302 @default.
- W3174917907 hasConcept C188441871 @default.
- W3174917907 hasConcept C41008148 @default.
- W3174917907 hasConcept C41895202 @default.
- W3174917907 hasConcept C50644808 @default.
- W3174917907 hasConcept C7149132 @default.
- W3174917907 hasConcept C95623464 @default.
- W3174917907 hasConceptScore W3174917907C113238511 @default.
- W3174917907 hasConceptScore W3174917907C119857082 @default.
- W3174917907 hasConceptScore W3174917907C138885662 @default.
- W3174917907 hasConceptScore W3174917907C153083717 @default.
- W3174917907 hasConceptScore W3174917907C153180895 @default.
- W3174917907 hasConceptScore W3174917907C154945302 @default.
- W3174917907 hasConceptScore W3174917907C188441871 @default.
- W3174917907 hasConceptScore W3174917907C41008148 @default.
- W3174917907 hasConceptScore W3174917907C41895202 @default.
- W3174917907 hasConceptScore W3174917907C50644808 @default.
- W3174917907 hasConceptScore W3174917907C7149132 @default.
- W3174917907 hasConceptScore W3174917907C95623464 @default.
- W3174917907 hasLocation W31749179071 @default.
- W3174917907 hasLocation W31749179072 @default.
- W3174917907 hasOpenAccess W3174917907 @default.
- W3174917907 hasPrimaryLocation W31749179071 @default.
- W3174917907 hasRelatedWork W2249953602 @default.
- W3174917907 hasRelatedWork W2947839263 @default.
- W3174917907 hasRelatedWork W2962876041 @default.
- W3174917907 hasRelatedWork W2980176872 @default.
- W3174917907 hasRelatedWork W3090555870 @default.
- W3174917907 hasRelatedWork W3107204728 @default.
- W3174917907 hasRelatedWork W3108503355 @default.
- W3174917907 hasRelatedWork W3120400911 @default.
- W3174917907 hasRelatedWork W4226420367 @default.
- W3174917907 hasRelatedWork W4287591324 @default.
- W3174917907 isParatext "false" @default.
- W3174917907 isRetracted "false" @default.
- W3174917907 magId "3174917907" @default.
- W3174917907 workType "article" @default.