FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3174039247> ?p ?o ?g. }

• W3174039247 abstract "For an image query, unsupervised contrastive learning labels crops of the same image as positives, and other image crops as negatives. Although intuitive, such a native label assignment strategy cannot reveal the underlying semantic similarity between a query and its positives and negatives, and impairs performance, since some negatives are semantically similar to the query or even share the same semantic class as the query. In this work, we first prove that for contrastive learning, inaccurate label assignment heavily impairs its generalization for semantic instance discrimination, while accurate labels benefit its generalization. Inspired by this theory, we propose a novel self-labeling refinement approach for contrastive learning. It improves the label quality via two complementary modules: (i) self-labeling refinery (SLR) to generate accurate labels and (ii) momentum mixup (MM) to enhance similarity between query and its positive. SLR uses a positive of a query to estimate semantic similarity between a query and its positive and negatives, and combines estimated similarity with vanilla label assignment in contrastive learning to iteratively generate more accurate and informative soft labels. We theoretically show that our SLR can exactly recover the true semantic labels of label-corrupted data, and supervises networks to achieve zero prediction error on classification tasks. MM randomly combines queries and positives to increase semantic similarity between the generated virtual queries and their positives so as to improves label accuracy. Experimental results on CIFAR10, ImageNet, VOC and COCO show the effectiveness of our method. PyTorch code and model will be released online." @default.
• W3174039247 created "2021-07-05" @default.
• W3174039247 creator A5010883708 @default.
• W3174039247 creator A5032046813 @default.
• W3174039247 creator A5074834854 @default.
• W3174039247 creator A5087073100 @default.
• W3174039247 date "2021-06-28" @default.
• W3174039247 modified "2023-09-27" @default.
• W3174039247 title "A Theory-Driven Self-Labeling Refinement Method for Contrastive Representation Learning" @default.
• W3174039247 cites W1786332878 @default.
• W3174039247 cites W1821462560 @default.
• W3174039247 cites W1861492603 @default.
• W3174039247 cites W2031489346 @default.
• W3174039247 cites W2108598243 @default.
• W3174039247 cites W2121056381 @default.
• W3174039247 cites W2138621090 @default.
• W3174039247 cites W2194775991 @default.
• W3174039247 cites W2321533354 @default.
• W3174039247 cites W2326925005 @default.
• W3174039247 cites W2610332124 @default.
• W3174039247 cites W2614119628 @default.
• W3174039247 cites W2626325961 @default.
• W3174039247 cites W2765407302 @default.
• W3174039247 cites W2798991696 @default.
• W3174039247 cites W2802198257 @default.
• W3174039247 cites W2842511635 @default.
• W3174039247 cites W2886067286 @default.
• W3174039247 cites W2894604724 @default.
• W3174039247 cites W2912811302 @default.
• W3174039247 cites W2917551568 @default.
• W3174039247 cites W2944828972 @default.
• W3174039247 cites W2948210185 @default.
• W3174039247 cites W2949517790 @default.
• W3174039247 cites W2951548721 @default.
• W3174039247 cites W2951873722 @default.
• W3174039247 cites W2952132225 @default.
• W3174039247 cites W2962698540 @default.
• W3174039247 cites W2962742544 @default.
• W3174039247 cites W2962742960 @default.
• W3174039247 cites W2962824366 @default.
• W3174039247 cites W2962845550 @default.
• W3174039247 cites W2962990163 @default.
• W3174039247 cites W2963097630 @default.
• W3174039247 cites W2963263347 @default.
• W3174039247 cites W2964098911 @default.
• W3174039247 cites W2964121744 @default.
• W3174039247 cites W2970241862 @default.
• W3174039247 cites W2971155163 @default.
• W3174039247 cites W2986405467 @default.
• W3174039247 cites W2998388430 @default.
• W3174039247 cites W3009561768 @default.
• W3174039247 cites W3018265077 @default.
• W3174039247 cites W3021189130 @default.
• W3174039247 cites W3026092005 @default.
• W3174039247 cites W3034781633 @default.
• W3174039247 cites W3034978746 @default.
• W3174039247 cites W3035524453 @default.
• W3174039247 cites W3037144731 @default.
• W3174039247 cites W3092113703 @default.
• W3174039247 cites W3093423309 @default.
• W3174039247 cites W3095121901 @default.
• W3174039247 cites W3099306795 @default.
• W3174039247 cites W3100859887 @default.
• W3174039247 cites W3101821705 @default.
• W3174039247 cites W3102363610 @default.
• W3174039247 cites W3107668149 @default.
• W3174039247 cites W3118608800 @default.
• W3174039247 cites W3122325173 @default.
• W3174039247 cites W3130223764 @default.
• W3174039247 cites W3131407161 @default.
• W3174039247 cites W3136810184 @default.
• W3174039247 doi "https://doi.org/10.48550/arxiv.2106.14749" @default.
• W3174039247 hasPublicationYear "2021" @default.
• W3174039247 type Work @default.
• W3174039247 sameAs 3174039247 @default.
• W3174039247 citedByCount "0" @default.
• W3174039247 crossrefType "posted-content" @default.
• W3174039247 hasAuthorship W3174039247A5010883708 @default.
• W3174039247 hasAuthorship W3174039247A5032046813 @default.
• W3174039247 hasAuthorship W3174039247A5074834854 @default.
• W3174039247 hasAuthorship W3174039247A5087073100 @default.
• W3174039247 hasBestOaLocation W31740392471 @default.
• W3174039247 hasConcept C103278499 @default.
• W3174039247 hasConcept C115961682 @default.
• W3174039247 hasConcept C119857082 @default.
• W3174039247 hasConcept C130318100 @default.
• W3174039247 hasConcept C134306372 @default.
• W3174039247 hasConcept C153180895 @default.
• W3174039247 hasConcept C154945302 @default.
• W3174039247 hasConcept C177148314 @default.
• W3174039247 hasConcept C204321447 @default.
• W3174039247 hasConcept C2777212361 @default.
• W3174039247 hasConcept C33923547 @default.
• W3174039247 hasConcept C41008148 @default.
• W3174039247 hasConcept C64869954 @default.
• W3174039247 hasConceptScore W3174039247C103278499 @default.
• W3174039247 hasConceptScore W3174039247C115961682 @default.
• W3174039247 hasConceptScore W3174039247C119857082 @default.
• W3174039247 hasConceptScore W3174039247C130318100 @default.
• W3174039247 hasConceptScore W3174039247C134306372 @default.

• next