FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2608395138> ?p ?o ?g. }

• W2608395138 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.
• W2608395138 created "2017-05-05" @default.
• W2608395138 creator A5061001839 @default.
• W2608395138 creator A5070594684 @default.
• W2608395138 date "2017-01-01" @default.
• W2608395138 modified "2023-09-25" @default.
• W2608395138 title "Unfolding and Shrinking Neural Machine Translation Ensembles" @default.
• W2608395138 cites W1534477342 @default.
• W2608395138 cites W1606347560 @default.
• W2608395138 cites W1821462560 @default.
• W2608395138 cites W2059174629 @default.
• W2608395138 cites W2113021982 @default.
• W2608395138 cites W2114766824 @default.
• W2608395138 cites W2125389748 @default.
• W2608395138 cites W2130942839 @default.
• W2608395138 cites W2135293965 @default.
• W2608395138 cites W2146502635 @default.
• W2608395138 cites W2157331557 @default.
• W2608395138 cites W2161591461 @default.
• W2608395138 cites W2167215970 @default.
• W2608395138 cites W2294059674 @default.
• W2608395138 cites W2294370754 @default.
• W2608395138 cites W2294543795 @default.
• W2608395138 cites W2312434537 @default.
• W2608395138 cites W2400065810 @default.
• W2608395138 cites W2507699225 @default.
• W2608395138 cites W2512924740 @default.
• W2608395138 cites W2549252813 @default.
• W2608395138 cites W2573197213 @default.
• W2608395138 cites W2576482813 @default.
• W2608395138 cites W2587694128 @default.
• W2608395138 cites W2915926444 @default.
• W2608395138 cites W2962784628 @default.
• W2608395138 cites W2962801832 @default.
• W2608395138 cites W2962867687 @default.
• W2608395138 cites W2962902089 @default.
• W2608395138 cites W2962907349 @default.
• W2608395138 cites W2963251942 @default.
• W2608395138 cites W2963643655 @default.
• W2608395138 cites W2963674932 @default.
• W2608395138 cites W2963991999 @default.
• W2608395138 cites W2964217848 @default.
• W2608395138 cites W2964308564 @default.
• W2608395138 cites W6908809 @default.
• W2608395138 doi "https://doi.org/10.18653/v1/d17-1208" @default.
• W2608395138 hasPublicationYear "2017" @default.
• W2608395138 type Work @default.
• W2608395138 sameAs 2608395138 @default.
• W2608395138 citedByCount "10" @default.
• W2608395138 countsByYear W26083951382018 @default.
• W2608395138 countsByYear W26083951382019 @default.
• W2608395138 countsByYear W26083951382020 @default.
• W2608395138 countsByYear W26083951382022 @default.
• W2608395138 crossrefType "proceedings-article" @default.
• W2608395138 hasAuthorship W2608395138A5061001839 @default.
• W2608395138 hasAuthorship W2608395138A5070594684 @default.
• W2608395138 hasBestOaLocation W26083951381 @default.
• W2608395138 hasConcept C104317684 @default.
• W2608395138 hasConcept C105580179 @default.
• W2608395138 hasConcept C111030470 @default.
• W2608395138 hasConcept C111472728 @default.
• W2608395138 hasConcept C11413529 @default.
• W2608395138 hasConcept C119857082 @default.
• W2608395138 hasConcept C138885662 @default.
• W2608395138 hasConcept C149364088 @default.
• W2608395138 hasConcept C154945302 @default.
• W2608395138 hasConcept C173608175 @default.
• W2608395138 hasConcept C177264268 @default.
• W2608395138 hasConcept C185592680 @default.
• W2608395138 hasConcept C187782996 @default.
• W2608395138 hasConcept C199360897 @default.
• W2608395138 hasConcept C203005215 @default.
• W2608395138 hasConcept C2779530757 @default.
• W2608395138 hasConcept C41008148 @default.
• W2608395138 hasConcept C50644808 @default.
• W2608395138 hasConcept C55493867 @default.
• W2608395138 hasConcept C57273362 @default.
• W2608395138 hasConcept C68339613 @default.
• W2608395138 hasConceptScore W2608395138C104317684 @default.
• W2608395138 hasConceptScore W2608395138C105580179 @default.
• W2608395138 hasConceptScore W2608395138C111030470 @default.
• W2608395138 hasConceptScore W2608395138C111472728 @default.
• W2608395138 hasConceptScore W2608395138C11413529 @default.
• W2608395138 hasConceptScore W2608395138C119857082 @default.
• W2608395138 hasConceptScore W2608395138C138885662 @default.
• W2608395138 hasConceptScore W2608395138C149364088 @default.
• W2608395138 hasConceptScore W2608395138C154945302 @default.
• W2608395138 hasConceptScore W2608395138C173608175 @default.
• W2608395138 hasConceptScore W2608395138C177264268 @default.
• W2608395138 hasConceptScore W2608395138C185592680 @default.
• W2608395138 hasConceptScore W2608395138C187782996 @default.
• W2608395138 hasConceptScore W2608395138C199360897 @default.
• W2608395138 hasConceptScore W2608395138C203005215 @default.
• W2608395138 hasConceptScore W2608395138C2779530757 @default.
• W2608395138 hasConceptScore W2608395138C41008148 @default.
• W2608395138 hasConceptScore W2608395138C50644808 @default.
• W2608395138 hasConceptScore W2608395138C55493867 @default.
• W2608395138 hasConceptScore W2608395138C57273362 @default.
• W2608395138 hasConceptScore W2608395138C68339613 @default.
• W2608395138 hasLocation W26083951381 @default.

• next