FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2950461903> ?p ?o ?g. }

• W2950461903 abstract "Ensembling is a well-known technique in neural machine translation (NMT) to improve system performance. Instead of a single neural net, multiple neural nets with the same topology are trained separately, and the decoder generates predictions by averaging over the individual models. Ensembling often improves the quality of the generated translations drastically. However, it is not suitable for production systems because it is cumbersome and slow. This work aims to reduce the runtime to be on par with a single system without compromising the translation quality. First, we show that the ensemble can be unfolded into a single large neural network which imitates the output of the ensemble system. We show that unfolding can already improve the runtime in practice since more work can be done on the GPU. We proceed by describing a set of techniques to shrink the unfolded network by reducing the dimensionality of layers. On Japanese-English we report that the resulting network has the size and decoding speed of a single NMT network but performs on the level of a 3-ensemble system." @default.
• W2950461903 created "2019-06-27" @default.
• W2950461903 creator A5061001839 @default.
• W2950461903 creator A5070594684 @default.
• W2950461903 date "2017-04-11" @default.
• W2950461903 modified "2023-09-27" @default.
• W2950461903 title "Unfolding and Shrinking Neural Machine Translation Ensembles" @default.
• W2950461903 cites W1534477342 @default.
• W2950461903 cites W1606347560 @default.
• W2950461903 cites W1821462560 @default.
• W2950461903 cites W2059174629 @default.
• W2950461903 cites W2113021982 @default.
• W2950461903 cites W2114766824 @default.
• W2950461903 cites W2114912785 @default.
• W2950461903 cites W2146502635 @default.
• W2950461903 cites W2167215970 @default.
• W2950461903 cites W2294370754 @default.
• W2950461903 cites W2294543795 @default.
• W2950461903 cites W2337978879 @default.
• W2950461903 cites W2507699225 @default.
• W2950461903 cites W2512924740 @default.
• W2950461903 cites W2525778437 @default.
• W2950461903 cites W2549252813 @default.
• W2950461903 cites W2573197213 @default.
• W2950461903 cites W2587694128 @default.
• W2950461903 cites W2949888546 @default.
• W2950461903 cites W2952533036 @default.
• W2950461903 cites W2952899695 @default.
• W2950461903 cites W2962784628 @default.
• W2950461903 cites W2962801832 @default.
• W2950461903 cites W2962902089 @default.
• W2950461903 cites W2963674932 @default.
• W2950461903 cites W2964308564 @default.
• W2950461903 cites W3037950864 @default.
• W2950461903 cites W6908809 @default.
• W2950461903 hasPublicationYear "2017" @default.
• W2950461903 type Work @default.
• W2950461903 sameAs 2950461903 @default.
• W2950461903 citedByCount "2" @default.
• W2950461903 countsByYear W29504619032018 @default.
• W2950461903 crossrefType "posted-content" @default.
• W2950461903 hasAuthorship W2950461903A5061001839 @default.
• W2950461903 hasAuthorship W2950461903A5070594684 @default.
• W2950461903 hasConcept C104317684 @default.
• W2950461903 hasConcept C105580179 @default.
• W2950461903 hasConcept C111030470 @default.
• W2950461903 hasConcept C111472728 @default.
• W2950461903 hasConcept C11413529 @default.
• W2950461903 hasConcept C119857082 @default.
• W2950461903 hasConcept C138885662 @default.
• W2950461903 hasConcept C149364088 @default.
• W2950461903 hasConcept C154945302 @default.
• W2950461903 hasConcept C177264268 @default.
• W2950461903 hasConcept C185592680 @default.
• W2950461903 hasConcept C187782996 @default.
• W2950461903 hasConcept C199360897 @default.
• W2950461903 hasConcept C203005215 @default.
• W2950461903 hasConcept C2779530757 @default.
• W2950461903 hasConcept C41008148 @default.
• W2950461903 hasConcept C50644808 @default.
• W2950461903 hasConcept C55493867 @default.
• W2950461903 hasConcept C57273362 @default.
• W2950461903 hasConceptScore W2950461903C104317684 @default.
• W2950461903 hasConceptScore W2950461903C105580179 @default.
• W2950461903 hasConceptScore W2950461903C111030470 @default.
• W2950461903 hasConceptScore W2950461903C111472728 @default.
• W2950461903 hasConceptScore W2950461903C11413529 @default.
• W2950461903 hasConceptScore W2950461903C119857082 @default.
• W2950461903 hasConceptScore W2950461903C138885662 @default.
• W2950461903 hasConceptScore W2950461903C149364088 @default.
• W2950461903 hasConceptScore W2950461903C154945302 @default.
• W2950461903 hasConceptScore W2950461903C177264268 @default.
• W2950461903 hasConceptScore W2950461903C185592680 @default.
• W2950461903 hasConceptScore W2950461903C187782996 @default.
• W2950461903 hasConceptScore W2950461903C199360897 @default.
• W2950461903 hasConceptScore W2950461903C203005215 @default.
• W2950461903 hasConceptScore W2950461903C2779530757 @default.
• W2950461903 hasConceptScore W2950461903C41008148 @default.
• W2950461903 hasConceptScore W2950461903C50644808 @default.
• W2950461903 hasConceptScore W2950461903C55493867 @default.
• W2950461903 hasConceptScore W2950461903C57273362 @default.
• W2950461903 hasLocation W29504619031 @default.
• W2950461903 hasOpenAccess W2950461903 @default.
• W2950461903 hasPrimaryLocation W29504619031 @default.
• W2950461903 hasRelatedWork W1525395049 @default.
• W2950461903 hasRelatedWork W1843283526 @default.
• W2950461903 hasRelatedWork W2181028625 @default.
• W2950461903 hasRelatedWork W2251077983 @default.
• W2950461903 hasRelatedWork W2289504355 @default.
• W2950461903 hasRelatedWork W2414666004 @default.
• W2950461903 hasRelatedWork W2507936800 @default.
• W2950461903 hasRelatedWork W2739542029 @default.
• W2950461903 hasRelatedWork W2785668315 @default.
• W2950461903 hasRelatedWork W2798937790 @default.
• W2950461903 hasRelatedWork W2807434070 @default.
• W2950461903 hasRelatedWork W2812009592 @default.
• W2950461903 hasRelatedWork W2894304603 @default.
• W2950461903 hasRelatedWork W2901104906 @default.
• W2950461903 hasRelatedWork W2921245078 @default.
• W2950461903 hasRelatedWork W2949920209 @default.

• next