Matches in SemOpenAlex for { <https://semopenalex.org/work/W2989771104> ?p ?o ?g. }
- W2989771104 abstract "The hubness problem widely exists in high-dimensional embedding space and is a fundamental source of error for cross-modal matching tasks. In this work, we study the emergence of hubs in Visual Semantic Embeddings (VSE) with application to text-image matching. We analyze the pros and cons of two widely adopted optimization objectives for training VSE and propose a novel hubness-aware loss function (HAL) that addresses previous methods' defects. Unlike (Faghri et al.2018) which simply takes the hardest sample within a mini-batch, HAL takes all samples into account, using both local and global statistics to scale up the weights of hubs. We experiment our method with various configurations of model architectures and datasets. The method exhibits exceptionally good robustness and brings consistent improvement on the task of text-image matching across all settings. Specifically, under the same model architectures as (Faghri et al. 2018) and (Lee at al. 2018), by switching only the learning objective, we report a maximum R@1improvement of 7.4% on MS-COCO and 8.3% on Flickr30k." @default.
- W2989771104 created "2019-12-05" @default.
- W2989771104 creator A5010804247 @default.
- W2989771104 creator A5016849647 @default.
- W2989771104 creator A5026154387 @default.
- W2989771104 creator A5058665519 @default.
- W2989771104 date "2019-11-22" @default.
- W2989771104 modified "2023-09-24" @default.
- W2989771104 title "HAL: Improved Text-Image Matching by Mitigating Visual Semantic Hubs" @default.
- W2989771104 cites W1527575280 @default.
- W2989771104 cites W1533861849 @default.
- W2989771104 cites W1686810756 @default.
- W2989771104 cites W1861492603 @default.
- W2989771104 cites W1905882502 @default.
- W2989771104 cites W1924770834 @default.
- W2989771104 cites W2096630704 @default.
- W2989771104 cites W2102765684 @default.
- W2989771104 cites W2108598243 @default.
- W2989771104 cites W2123024445 @default.
- W2989771104 cites W2129219632 @default.
- W2989771104 cites W2135442311 @default.
- W2989771104 cites W2144935315 @default.
- W2989771104 cites W2156365280 @default.
- W2989771104 cites W2185175083 @default.
- W2989771104 cites W2194775991 @default.
- W2989771104 cites W2250646737 @default.
- W2989771104 cites W2252211741 @default.
- W2989771104 cites W2476034201 @default.
- W2989771104 cites W2552383788 @default.
- W2989771104 cites W2552579943 @default.
- W2989771104 cites W2560014971 @default.
- W2989771104 cites W2594021297 @default.
- W2989771104 cites W2603705233 @default.
- W2989771104 cites W2606473278 @default.
- W2989771104 cites W2613718673 @default.
- W2989771104 cites W2778100917 @default.
- W2989771104 cites W2798280084 @default.
- W2989771104 cites W2798782720 @default.
- W2989771104 cites W2810689564 @default.
- W2989771104 cites W2885201931 @default.
- W2989771104 cites W2891896107 @default.
- W2989771104 cites W2936420065 @default.
- W2989771104 cites W2949461431 @default.
- W2989771104 cites W2962964995 @default.
- W2989771104 cites W2963026686 @default.
- W2989771104 cites W2963061446 @default.
- W2989771104 cites W2963118869 @default.
- W2989771104 cites W2963350250 @default.
- W2989771104 cites W2963419157 @default.
- W2989771104 cites W2963499204 @default.
- W2989771104 cites W2963899908 @default.
- W2989771104 cites W2964271799 @default.
- W2989771104 cites W2964350391 @default.
- W2989771104 cites W2982260276 @default.
- W2989771104 cites W3099206234 @default.
- W2989771104 cites W3104379732 @default.
- W2989771104 cites W68733909 @default.
- W2989771104 cites W92662927 @default.
- W2989771104 doi "https://doi.org/10.48550/arxiv.1911.10097" @default.
- W2989771104 hasPublicationYear "2019" @default.
- W2989771104 type Work @default.
- W2989771104 sameAs 2989771104 @default.
- W2989771104 citedByCount "0" @default.
- W2989771104 crossrefType "posted-content" @default.
- W2989771104 hasAuthorship W2989771104A5010804247 @default.
- W2989771104 hasAuthorship W2989771104A5016849647 @default.
- W2989771104 hasAuthorship W2989771104A5026154387 @default.
- W2989771104 hasAuthorship W2989771104A5058665519 @default.
- W2989771104 hasBestOaLocation W29897711041 @default.
- W2989771104 hasConcept C104317684 @default.
- W2989771104 hasConcept C105795698 @default.
- W2989771104 hasConcept C115961682 @default.
- W2989771104 hasConcept C119857082 @default.
- W2989771104 hasConcept C153180895 @default.
- W2989771104 hasConcept C154945302 @default.
- W2989771104 hasConcept C162324750 @default.
- W2989771104 hasConcept C165064840 @default.
- W2989771104 hasConcept C185592680 @default.
- W2989771104 hasConcept C187736073 @default.
- W2989771104 hasConcept C188027245 @default.
- W2989771104 hasConcept C2778493491 @default.
- W2989771104 hasConcept C2780451532 @default.
- W2989771104 hasConcept C33923547 @default.
- W2989771104 hasConcept C41008148 @default.
- W2989771104 hasConcept C41608201 @default.
- W2989771104 hasConcept C55493867 @default.
- W2989771104 hasConcept C63479239 @default.
- W2989771104 hasConcept C71139939 @default.
- W2989771104 hasConceptScore W2989771104C104317684 @default.
- W2989771104 hasConceptScore W2989771104C105795698 @default.
- W2989771104 hasConceptScore W2989771104C115961682 @default.
- W2989771104 hasConceptScore W2989771104C119857082 @default.
- W2989771104 hasConceptScore W2989771104C153180895 @default.
- W2989771104 hasConceptScore W2989771104C154945302 @default.
- W2989771104 hasConceptScore W2989771104C162324750 @default.
- W2989771104 hasConceptScore W2989771104C165064840 @default.
- W2989771104 hasConceptScore W2989771104C185592680 @default.
- W2989771104 hasConceptScore W2989771104C187736073 @default.
- W2989771104 hasConceptScore W2989771104C188027245 @default.
- W2989771104 hasConceptScore W2989771104C2778493491 @default.