FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2972175708> ?p ?o ?g. }

• W2972175708 abstract "Neural models for NLP typically use large numbers of parameters to reach state-of-the-art performance, which can lead to excessive memory usage and increased runtime. We present a structure learning method for learning sparse, parameter-efficient NLP models. Our method applies group lasso to rational RNNs (Peng et al., 2018), a family of models that is closely connected to weighted finite-state automata (WFSAs). We take advantage of rational RNNs' natural grouping of the weights, so the group lasso penalty directly removes WFSA states, substantially reducing the number of parameters in the model. Our experiments on a number of sentiment analysis datasets, using both GloVe and BERT embeddings, show that our approach learns neural structures which have fewer parameters without sacrificing performance relative to parameter-rich baselines. Our method also highlights the interpretable properties of rational RNNs. We show that sparsifying such models makes them easier to visualize, and we present models that rely exclusively on as few as three WFSAs after pruning more than 90% of the weights. We publicly release our code." @default.
• W2972175708 created "2019-09-12" @default.
• W2972175708 creator A5007903277 @default.
• W2972175708 creator A5008013895 @default.
• W2972175708 creator A5079854850 @default.
• W2972175708 creator A5088517824 @default.
• W2972175708 date "2019-09-06" @default.
• W2972175708 modified "2023-10-01" @default.
• W2972175708 title "RNN Architecture Learning with Sparse Regularization" @default.
• W2972175708 cites W138607541 @default.
• W2972175708 cites W17986227 @default.
• W2972175708 cites W2007321142 @default.
• W2972175708 cites W2064675550 @default.
• W2972175708 cites W2066539191 @default.
• W2972175708 cites W2114766824 @default.
• W2972175708 cites W2138019504 @default.
• W2972175708 cites W2157331557 @default.
• W2972175708 cites W2162157640 @default.
• W2972175708 cites W2163302275 @default.
• W2972175708 cites W2250539671 @default.
• W2972175708 cites W2460144244 @default.
• W2972175708 cites W2553397501 @default.
• W2972175708 cites W2556623528 @default.
• W2972175708 cites W2559554776 @default.
• W2972175708 cites W2751185861 @default.
• W2972175708 cites W2810075754 @default.
• W2972175708 cites W2889175957 @default.
• W2972175708 cites W2890177507 @default.
• W2972175708 cites W2913535645 @default.
• W2972175708 cites W2963000224 @default.
• W2972175708 cites W2963247446 @default.
• W2972175708 cites W2963341956 @default.
• W2972175708 cites W2963374479 @default.
• W2972175708 cites W2963759574 @default.
• W2972175708 cites W2963809228 @default.
• W2972175708 cites W2963813662 @default.
• W2972175708 cites W2963841132 @default.
• W2972175708 cites W2963921497 @default.
• W2972175708 cites W2964121744 @default.
• W2972175708 cites W2964217527 @default.
• W2972175708 cites W2964308564 @default.
• W2972175708 cites W581956982 @default.
• W2972175708 hasPublicationYear "2019" @default.
• W2972175708 type Work @default.
• W2972175708 sameAs 2972175708 @default.
• W2972175708 citedByCount "0" @default.
• W2972175708 crossrefType "posted-content" @default.
• W2972175708 hasAuthorship W2972175708A5007903277 @default.
• W2972175708 hasAuthorship W2972175708A5008013895 @default.
• W2972175708 hasAuthorship W2972175708A5079854850 @default.
• W2972175708 hasAuthorship W2972175708A5088517824 @default.
• W2972175708 hasConcept C108010975 @default.
• W2972175708 hasConcept C119857082 @default.
• W2972175708 hasConcept C123657996 @default.
• W2972175708 hasConcept C136764020 @default.
• W2972175708 hasConcept C142362112 @default.
• W2972175708 hasConcept C147168706 @default.
• W2972175708 hasConcept C153349607 @default.
• W2972175708 hasConcept C154945302 @default.
• W2972175708 hasConcept C177264268 @default.
• W2972175708 hasConcept C199360897 @default.
• W2972175708 hasConcept C2776135515 @default.
• W2972175708 hasConcept C2776760102 @default.
• W2972175708 hasConcept C37616216 @default.
• W2972175708 hasConcept C41008148 @default.
• W2972175708 hasConcept C50644808 @default.
• W2972175708 hasConcept C6557445 @default.
• W2972175708 hasConcept C86803240 @default.
• W2972175708 hasConceptScore W2972175708C108010975 @default.
• W2972175708 hasConceptScore W2972175708C119857082 @default.
• W2972175708 hasConceptScore W2972175708C123657996 @default.
• W2972175708 hasConceptScore W2972175708C136764020 @default.
• W2972175708 hasConceptScore W2972175708C142362112 @default.
• W2972175708 hasConceptScore W2972175708C147168706 @default.
• W2972175708 hasConceptScore W2972175708C153349607 @default.
• W2972175708 hasConceptScore W2972175708C154945302 @default.
• W2972175708 hasConceptScore W2972175708C177264268 @default.
• W2972175708 hasConceptScore W2972175708C199360897 @default.
• W2972175708 hasConceptScore W2972175708C2776135515 @default.
• W2972175708 hasConceptScore W2972175708C2776760102 @default.
• W2972175708 hasConceptScore W2972175708C37616216 @default.
• W2972175708 hasConceptScore W2972175708C41008148 @default.
• W2972175708 hasConceptScore W2972175708C50644808 @default.
• W2972175708 hasConceptScore W2972175708C6557445 @default.
• W2972175708 hasConceptScore W2972175708C86803240 @default.
• W2972175708 hasLocation W29721757081 @default.
• W2972175708 hasOpenAccess W2972175708 @default.
• W2972175708 hasPrimaryLocation W29721757081 @default.
• W2972175708 hasRelatedWork W1761606332 @default.
• W2972175708 hasRelatedWork W202196065 @default.
• W2972175708 hasRelatedWork W2059554529 @default.
• W2972175708 hasRelatedWork W2576410866 @default.
• W2972175708 hasRelatedWork W2736296425 @default.
• W2972175708 hasRelatedWork W2761394014 @default.
• W2972175708 hasRelatedWork W2908692794 @default.
• W2972175708 hasRelatedWork W2912119436 @default.
• W2972175708 hasRelatedWork W2922440410 @default.
• W2972175708 hasRelatedWork W2952475472 @default.
• W2972175708 hasRelatedWork W3000610597 @default.
• W2972175708 hasRelatedWork W3004752032 @default.

• next