FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2957874408> ?p ?o ?g. }

• W2957874408 abstract "Recent convolutional neural networks (CNNs) have led to impressive performance but often suffer from poor calibration. They tend to be overconfident, with the model confidence not always reflecting the underlying true ambiguity and hardness. In this paper, we propose angular visual hardness (AVH), a score given by the normalized angular distance between the sample feature embedding and the target classifier to measure sample hardness. We validate this score with an in-depth and extensive scientific study, and observe that CNN models with the highest accuracy also have the best AVH scores. This agrees with an earlier finding that state-of-art models improve on the classification of harder examples. We observe that the training dynamics of AVH is vastly different compared to the training loss. Specifically, AVH quickly reaches a plateau for all samples even though the training loss keeps improving. This suggests the need for designing better loss functions that can target harder examples more effectively. We also find that AVH has a statistically significant correlation with human visual hardness. Finally, we demonstrate the benefit of AVH to a variety of applications such as self-training for domain adaptation and domain generalization." @default.
• W2957874408 created "2019-07-23" @default.
• W2957874408 creator A5014498545 @default.
• W2957874408 creator A5024993683 @default.
• W2957874408 creator A5030464101 @default.
• W2957874408 creator A5031842648 @default.
• W2957874408 creator A5054586129 @default.
• W2957874408 creator A5056503617 @default.
• W2957874408 creator A5061193324 @default.
• W2957874408 date "2019-12-04" @default.
• W2957874408 modified "2023-09-25" @default.
• W2957874408 title "Angular Visual Hardness" @default.
• W2957874408 cites W1509966554 @default.
• W2957874408 cites W1583975023 @default.
• W2957874408 cites W1686810756 @default.
• W2957874408 cites W1731081199 @default.
• W2957874408 cites W1821462560 @default.
• W2957874408 cites W1903029394 @default.
• W2957874408 cites W1995341919 @default.
• W2957874408 cites W1998808035 @default.
• W2957874408 cites W2098580305 @default.
• W2957874408 cites W2101926813 @default.
• W2957874408 cites W2102605133 @default.
• W2957874408 cites W2104094955 @default.
• W2957874408 cites W2107176174 @default.
• W2957874408 cites W2108598243 @default.
• W2957874408 cites W2111959010 @default.
• W2957874408 cites W2118045473 @default.
• W2957874408 cites W2132984949 @default.
• W2957874408 cites W2133665775 @default.
• W2957874408 cites W2134670479 @default.
• W2957874408 cites W2138621090 @default.
• W2957874408 cites W2145494108 @default.
• W2957874408 cites W2159291411 @default.
• W2957874408 cites W2166206801 @default.
• W2957874408 cites W2183646625 @default.
• W2957874408 cites W2194775991 @default.
• W2957874408 cites W2296073425 @default.
• W2957874408 cites W2328322609 @default.
• W2957874408 cites W2475327816 @default.
• W2957874408 cites W2555897561 @default.
• W2957874408 cites W2556372419 @default.
• W2957874408 cites W2573862116 @default.
• W2957874408 cites W2600383743 @default.
• W2957874408 cites W2612573399 @default.
• W2957874408 cites W2751592082 @default.
• W2957874408 cites W2763549966 @default.
• W2957874408 cites W2763767712 @default.
• W2957874408 cites W2766897166 @default.
• W2957874408 cites W2767015886 @default.
• W2957874408 cites W2771773135 @default.
• W2957874408 cites W2786559811 @default.
• W2957874408 cites W2788481061 @default.
• W2957874408 cites W2794302998 @default.
• W2957874408 cites W2804697534 @default.
• W2957874408 cites W2888339491 @default.
• W2957874408 cites W2889965839 @default.
• W2957874408 cites W2895779420 @default.
• W2957874408 cites W2903365351 @default.
• W2957874408 cites W2911742574 @default.
• W2957874408 cites W2912512139 @default.
• W2957874408 cites W2950984716 @default.
• W2957874408 cites W2952610664 @default.
• W2957874408 cites W2952848538 @default.
• W2957874408 cites W2953274220 @default.
• W2957874408 cites W2953937638 @default.
• W2957874408 cites W2962785568 @default.
• W2957874408 cites W2962898354 @default.
• W2957874408 cites W2963026686 @default.
• W2957874408 cites W2963078860 @default.
• W2957874408 cites W2963207607 @default.
• W2957874408 cites W2963224870 @default.
• W2957874408 cites W2963325056 @default.
• W2957874408 cites W2963350250 @default.
• W2957874408 cites W2963446712 @default.
• W2957874408 cites W2963466847 @default.
• W2957874408 cites W2963656735 @default.
• W2957874408 cites W2964059111 @default.
• W2957874408 cites W2964139811 @default.
• W2957874408 cites W2964155802 @default.
• W2957874408 cites W2964212410 @default.
• W2957874408 cites W2964237080 @default.
• W2957874408 cites W2969985801 @default.
• W2957874408 cites W2971228020 @default.
• W2957874408 cites W2979484367 @default.
• W2957874408 cites W2985406498 @default.
• W2957874408 cites W3015548926 @default.
• W2957874408 cites W3034938706 @default.
• W2957874408 cites W3037590790 @default.
• W2957874408 cites W3098596645 @default.
• W2957874408 cites W3101227480 @default.
• W2957874408 cites W2600537992 @default.
• W2957874408 hasPublicationYear "2019" @default.
• W2957874408 type Work @default.
• W2957874408 sameAs 2957874408 @default.
• W2957874408 citedByCount "2" @default.
• W2957874408 countsByYear W29578744082020 @default.
• W2957874408 crossrefType "posted-content" @default.
• W2957874408 hasAuthorship W2957874408A5014498545 @default.
• W2957874408 hasAuthorship W2957874408A5024993683 @default.

• next