Matches in SemOpenAlex for { <https://semopenalex.org/work/W2804069376> ?p ?o ?g. }
- W2804069376 abstract "We introduce hyperbolic attention networks to endow neural networks with enough capacity to match the complexity of data with hierarchical and power-law structure. A few recent approaches have successfully demonstrated the benefits of imposing hyperbolic geometry on the parameters of shallow networks. We extend this line of work by imposing hyperbolic geometry on the activations of neural networks. This allows us to exploit hyperbolic geometry to reason about embeddings produced by deep networks. We achieve this by re-expressing the ubiquitous mechanism of soft attention in terms of operations defined for hyperboloid and Klein models. Our method shows improvements in terms of generalization on neural machine translation, learning on graphs and visual question answering tasks while keeping the neural representations compact." @default.
- W2804069376 created "2018-06-01" @default.
- W2804069376 creator A5001284503 @default.
- W2804069376 creator A5008260665 @default.
- W2804069376 creator A5025389620 @default.
- W2804069376 creator A5031378020 @default.
- W2804069376 creator A5041145688 @default.
- W2804069376 creator A5041745766 @default.
- W2804069376 creator A5043910056 @default.
- W2804069376 creator A5059792149 @default.
- W2804069376 creator A5061984716 @default.
- W2804069376 creator A5082304130 @default.
- W2804069376 creator A5089917436 @default.
- W2804069376 date "2018-05-24" @default.
- W2804069376 modified "2023-10-18" @default.
- W2804069376 title "Hyperbolic Attention Networks" @default.
- W2804069376 cites W1662382123 @default.
- W2804069376 cites W1862455499 @default.
- W2804069376 cites W1968625547 @default.
- W2804069376 cites W2001070843 @default.
- W2804069376 cites W2048176942 @default.
- W2804069376 cites W2052629646 @default.
- W2804069376 cites W2133564696 @default.
- W2804069376 cites W2145339207 @default.
- W2804069376 cites W2147438222 @default.
- W2804069376 cites W2199357291 @default.
- W2804069376 cites W2244807774 @default.
- W2804069376 cites W2401915092 @default.
- W2804069376 cites W2519887557 @default.
- W2804069376 cites W2561715562 @default.
- W2804069376 cites W2617943363 @default.
- W2804069376 cites W2624614404 @default.
- W2804069376 cites W2626778328 @default.
- W2804069376 cites W2766453196 @default.
- W2804069376 cites W2770201307 @default.
- W2804069376 cites W2787337315 @default.
- W2804069376 cites W2789541106 @default.
- W2804069376 cites W2950627632 @default.
- W2804069376 cites W2951619830 @default.
- W2804069376 cites W2951846985 @default.
- W2804069376 cites W2952254971 @default.
- W2804069376 cites W2962749469 @default.
- W2804069376 cites W2964321699 @default.
- W2804069376 cites W3102543255 @default.
- W2804069376 cites W638895084 @default.
- W2804069376 hasPublicationYear "2018" @default.
- W2804069376 type Work @default.
- W2804069376 sameAs 2804069376 @default.
- W2804069376 citedByCount "33" @default.
- W2804069376 countsByYear W28040693762018 @default.
- W2804069376 countsByYear W28040693762019 @default.
- W2804069376 countsByYear W28040693762020 @default.
- W2804069376 countsByYear W28040693762021 @default.
- W2804069376 crossrefType "posted-content" @default.
- W2804069376 hasAuthorship W2804069376A5001284503 @default.
- W2804069376 hasAuthorship W2804069376A5008260665 @default.
- W2804069376 hasAuthorship W2804069376A5025389620 @default.
- W2804069376 hasAuthorship W2804069376A5031378020 @default.
- W2804069376 hasAuthorship W2804069376A5041145688 @default.
- W2804069376 hasAuthorship W2804069376A5041745766 @default.
- W2804069376 hasAuthorship W2804069376A5043910056 @default.
- W2804069376 hasAuthorship W2804069376A5059792149 @default.
- W2804069376 hasAuthorship W2804069376A5061984716 @default.
- W2804069376 hasAuthorship W2804069376A5082304130 @default.
- W2804069376 hasAuthorship W2804069376A5089917436 @default.
- W2804069376 hasConcept C114614502 @default.
- W2804069376 hasConcept C134306372 @default.
- W2804069376 hasConcept C135796866 @default.
- W2804069376 hasConcept C147168706 @default.
- W2804069376 hasConcept C154945302 @default.
- W2804069376 hasConcept C165696696 @default.
- W2804069376 hasConcept C177148314 @default.
- W2804069376 hasConcept C184720557 @default.
- W2804069376 hasConcept C206352148 @default.
- W2804069376 hasConcept C2524010 @default.
- W2804069376 hasConcept C33923547 @default.
- W2804069376 hasConcept C38652104 @default.
- W2804069376 hasConcept C41008148 @default.
- W2804069376 hasConcept C50644808 @default.
- W2804069376 hasConcept C68363185 @default.
- W2804069376 hasConcept C80444323 @default.
- W2804069376 hasConcept C99564226 @default.
- W2804069376 hasConceptScore W2804069376C114614502 @default.
- W2804069376 hasConceptScore W2804069376C134306372 @default.
- W2804069376 hasConceptScore W2804069376C135796866 @default.
- W2804069376 hasConceptScore W2804069376C147168706 @default.
- W2804069376 hasConceptScore W2804069376C154945302 @default.
- W2804069376 hasConceptScore W2804069376C165696696 @default.
- W2804069376 hasConceptScore W2804069376C177148314 @default.
- W2804069376 hasConceptScore W2804069376C184720557 @default.
- W2804069376 hasConceptScore W2804069376C206352148 @default.
- W2804069376 hasConceptScore W2804069376C2524010 @default.
- W2804069376 hasConceptScore W2804069376C33923547 @default.
- W2804069376 hasConceptScore W2804069376C38652104 @default.
- W2804069376 hasConceptScore W2804069376C41008148 @default.
- W2804069376 hasConceptScore W2804069376C50644808 @default.
- W2804069376 hasConceptScore W2804069376C68363185 @default.
- W2804069376 hasConceptScore W2804069376C80444323 @default.
- W2804069376 hasConceptScore W2804069376C99564226 @default.
- W2804069376 hasLocation W28040693761 @default.