FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3026092005> ?p ?o ?g. }

• W3026092005 abstract "Contrastive learning between multiple views of the data has recently achieved state of the art performance in the field of self-supervised representation learning. Despite its success, the influence of different view choices has been less studied. In this paper, we use empirical analysis to better understand the importance of view selection, and argue that we should reduce the mutual information (MI) between views while keeping task-relevant information intact. To verify this hypothesis, we devise unsupervised and semi-supervised frameworks that learn effective views by aiming to reduce their MI. We also consider data augmentation as a way to reduce MI, and show that increasing data augmentation indeed leads to decreasing MI and improves downstream classification accuracy. As a by-product, we also achieve a new state-of-the-art accuracy on unsupervised pre-training for ImageNet classification ($73%$ top-1 linear readoff with a ResNet-50). In addition, transferring our models to PASCAL VOC object detection and COCO instance segmentation consistently outperforms supervised pre-training. Code:this http URL" @default.
• W3026092005 created "2020-05-29" @default.
• W3026092005 creator A5004918186 @default.
• W3026092005 creator A5017456911 @default.
• W3026092005 creator A5045217258 @default.
• W3026092005 creator A5059925606 @default.
• W3026092005 creator A5060145891 @default.
• W3026092005 creator A5085958820 @default.
• W3026092005 date "2020-05-20" @default.
• W3026092005 modified "2023-09-27" @default.
• W3026092005 title "What Makes for Good Views for Contrastive Learning" @default.
• W3026092005 cites W125693051 @default.
• W3026092005 cites W1522301498 @default.
• W3026092005 cites W1577841048 @default.
• W3026092005 cites W1583912456 @default.
• W3026092005 cites W1686946872 @default.
• W3026092005 cites W1861492603 @default.
• W3026092005 cites W2031489346 @default.
• W3026092005 cites W2064675550 @default.
• W3026092005 cites W2099471712 @default.
• W3026092005 cites W2100115174 @default.
• W3026092005 cites W2108598243 @default.
• W3026092005 cites W2116435618 @default.
• W3026092005 cites W2118858186 @default.
• W3026092005 cites W2122925692 @default.
• W3026092005 cites W2123469175 @default.
• W3026092005 cites W2148349024 @default.
• W3026092005 cites W2153579005 @default.
• W3026092005 cites W2175760607 @default.
• W3026092005 cites W2194775991 @default.
• W3026092005 cites W2321533354 @default.
• W3026092005 cites W2326925005 @default.
• W3026092005 cites W2409550820 @default.
• W3026092005 cites W2490570449 @default.
• W3026092005 cites W2549139847 @default.
• W3026092005 cites W2555897561 @default.
• W3026092005 cites W2557579533 @default.
• W3026092005 cites W2558661413 @default.
• W3026092005 cites W2565639579 @default.
• W3026092005 cites W2599765304 @default.
• W3026092005 cites W2613718673 @default.
• W3026092005 cites W2741137940 @default.
• W3026092005 cites W2783047733 @default.
• W3026092005 cites W2798991696 @default.
• W3026092005 cites W2842511635 @default.
• W3026092005 cites W2885201931 @default.
• W3026092005 cites W2887997457 @default.
• W3026092005 cites W2900148439 @default.
• W3026092005 cites W2913939497 @default.
• W3026092005 cites W2935908327 @default.
• W3026092005 cites W2944828972 @default.
• W3026092005 cites W2946845979 @default.
• W3026092005 cites W2946856970 @default.
• W3026092005 cites W2948012107 @default.
• W3026092005 cites W2949178656 @default.
• W3026092005 cites W2949517790 @default.
• W3026092005 cites W2949759968 @default.
• W3026092005 cites W2950864153 @default.
• W3026092005 cites W2951004968 @default.
• W3026092005 cites W2962742544 @default.
• W3026092005 cites W2963016543 @default.
• W3026092005 cites W2963115079 @default.
• W3026092005 cites W2963139417 @default.
• W3026092005 cites W2963420272 @default.
• W3026092005 cites W2963644595 @default.
• W3026092005 cites W2964160928 @default.
• W3026092005 cites W2964241181 @default.
• W3026092005 cites W2964420626 @default.
• W3026092005 cites W2970241862 @default.
• W3026092005 cites W2970597249 @default.
• W3026092005 cites W2975357369 @default.
• W3026092005 cites W2975761646 @default.
• W3026092005 cites W2979579363 @default.
• W3026092005 cites W2981952041 @default.
• W3026092005 cites W2982376094 @default.
• W3026092005 cites W2987283559 @default.
• W3026092005 cites W2993466051 @default.
• W3026092005 cites W2994268455 @default.
• W3026092005 cites W2994536315 @default.
• W3026092005 cites W2995040292 @default.
• W3026092005 cites W2995181141 @default.
• W3026092005 cites W2995489995 @default.
• W3026092005 cites W2995607862 @default.
• W3026092005 cites W2999004587 @default.
• W3026092005 cites W3005680577 @default.
• W3026092005 cites W3007373432 @default.
• W3026092005 cites W3009561768 @default.
• W3026092005 cites W3010094231 @default.
• W3026092005 cites W3010874390 @default.
• W3026092005 cites W3012540512 @default.
• W3026092005 cites W3020933450 @default.
• W3026092005 cites W3083978735 @default.
• W3026092005 cites W3100345210 @default.
• W3026092005 cites W3115293622 @default.
• W3026092005 cites W3175156984 @default.
• W3026092005 cites W3189092450 @default.
• W3026092005 cites W343636949 @default.
• W3026092005 cites W2996163128 @default.
• W3026092005 hasPublicationYear "2020" @default.
• W3026092005 type Work @default.

• next