Matches in SemOpenAlex for { <https://semopenalex.org/work/W2980596743> ?p ?o ?g. }
- W2980596743 abstract "Neural sequence-to-sequence models, particularly the Transformer, are the state of the art in machine translation. Yet these neural networks are very sensitive to architecture and hyperparameter settings. Optimizing these settings by grid or random search is computationally expensive because it requires many training runs. In this paper, we incorporate architecture search into a single training run through auto-sizing, which uses regularization to delete neurons in a network over the course of training. On very low-resource language pairs, we show that auto-sizing can improve BLEU scores by up to 3.9 points while removing one-third of the parameters from the model." @default.
- W2980596743 created "2019-10-25" @default.
- W2980596743 creator A5005840801 @default.
- W2980596743 creator A5036026526 @default.
- W2980596743 creator A5047396367 @default.
- W2980596743 creator A5081708301 @default.
- W2980596743 creator A5091513740 @default.
- W2980596743 date "2019-10-01" @default.
- W2980596743 modified "2023-09-27" @default.
- W2980596743 title "Auto-Sizing the Transformer Network: Improving Speed, Efficiency, and Performance for Low-Resource Machine Translation" @default.
- W2980596743 cites W1928693685 @default.
- W2980596743 cites W1978259121 @default.
- W2980596743 cites W1986490585 @default.
- W2980596743 cites W2047092297 @default.
- W2980596743 cites W2073515924 @default.
- W2980596743 cites W2097998348 @default.
- W2980596743 cites W2098368939 @default.
- W2980596743 cites W2101105183 @default.
- W2980596743 cites W2106411961 @default.
- W2980596743 cites W2124807415 @default.
- W2980596743 cites W2130942839 @default.
- W2980596743 cites W2131241448 @default.
- W2980596743 cites W2143462372 @default.
- W2980596743 cites W2164278908 @default.
- W2980596743 cites W2184135559 @default.
- W2980596743 cites W2395935897 @default.
- W2980596743 cites W2410082850 @default.
- W2980596743 cites W2594990650 @default.
- W2980596743 cites W2909951054 @default.
- W2980596743 cites W2913535645 @default.
- W2980596743 cites W2917028965 @default.
- W2980596743 cites W2962784628 @default.
- W2980596743 cites W2963181996 @default.
- W2980596743 cites W2963233958 @default.
- W2980596743 cites W2963347649 @default.
- W2980596743 cites W2963403868 @default.
- W2980596743 cites W2963809228 @default.
- W2980596743 cites W2964121744 @default.
- W2980596743 cites W2964265128 @default.
- W2980596743 cites W2964308564 @default.
- W2980596743 hasPublicationYear "2019" @default.
- W2980596743 type Work @default.
- W2980596743 sameAs 2980596743 @default.
- W2980596743 citedByCount "1" @default.
- W2980596743 countsByYear W29805967432020 @default.
- W2980596743 crossrefType "posted-content" @default.
- W2980596743 hasAuthorship W2980596743A5005840801 @default.
- W2980596743 hasAuthorship W2980596743A5036026526 @default.
- W2980596743 hasAuthorship W2980596743A5047396367 @default.
- W2980596743 hasAuthorship W2980596743A5081708301 @default.
- W2980596743 hasAuthorship W2980596743A5091513740 @default.
- W2980596743 hasConcept C119599485 @default.
- W2980596743 hasConcept C119857082 @default.
- W2980596743 hasConcept C123657996 @default.
- W2980596743 hasConcept C127413603 @default.
- W2980596743 hasConcept C137293760 @default.
- W2980596743 hasConcept C142362112 @default.
- W2980596743 hasConcept C153349607 @default.
- W2980596743 hasConcept C154945302 @default.
- W2980596743 hasConcept C165801399 @default.
- W2980596743 hasConcept C173608175 @default.
- W2980596743 hasConcept C187691185 @default.
- W2980596743 hasConcept C203005215 @default.
- W2980596743 hasConcept C2524010 @default.
- W2980596743 hasConcept C2777767291 @default.
- W2980596743 hasConcept C33923547 @default.
- W2980596743 hasConcept C41008148 @default.
- W2980596743 hasConcept C50644808 @default.
- W2980596743 hasConcept C66322947 @default.
- W2980596743 hasConcept C68339613 @default.
- W2980596743 hasConcept C8642999 @default.
- W2980596743 hasConceptScore W2980596743C119599485 @default.
- W2980596743 hasConceptScore W2980596743C119857082 @default.
- W2980596743 hasConceptScore W2980596743C123657996 @default.
- W2980596743 hasConceptScore W2980596743C127413603 @default.
- W2980596743 hasConceptScore W2980596743C137293760 @default.
- W2980596743 hasConceptScore W2980596743C142362112 @default.
- W2980596743 hasConceptScore W2980596743C153349607 @default.
- W2980596743 hasConceptScore W2980596743C154945302 @default.
- W2980596743 hasConceptScore W2980596743C165801399 @default.
- W2980596743 hasConceptScore W2980596743C173608175 @default.
- W2980596743 hasConceptScore W2980596743C187691185 @default.
- W2980596743 hasConceptScore W2980596743C203005215 @default.
- W2980596743 hasConceptScore W2980596743C2524010 @default.
- W2980596743 hasConceptScore W2980596743C2777767291 @default.
- W2980596743 hasConceptScore W2980596743C33923547 @default.
- W2980596743 hasConceptScore W2980596743C41008148 @default.
- W2980596743 hasConceptScore W2980596743C50644808 @default.
- W2980596743 hasConceptScore W2980596743C66322947 @default.
- W2980596743 hasConceptScore W2980596743C68339613 @default.
- W2980596743 hasConceptScore W2980596743C8642999 @default.
- W2980596743 hasLocation W29805967431 @default.
- W2980596743 hasOpenAccess W2980596743 @default.
- W2980596743 hasPrimaryLocation W29805967431 @default.
- W2980596743 hasRelatedWork W1578315993 @default.
- W2980596743 hasRelatedWork W2799290373 @default.
- W2980596743 hasRelatedWork W2901508112 @default.
- W2980596743 hasRelatedWork W2903852246 @default.
- W2980596743 hasRelatedWork W2911072618 @default.
- W2980596743 hasRelatedWork W2915236044 @default.