FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2940446723> ?p ?o ?g. }

• W2940446723 abstract "Recurrent neural network language models (RNNLMs) have become an increasing popular choice for state-of-the-art speech recognition systems. RNNLMs are normally trained by minimizing the cross entropy (CE) using the stochastic gradient descent (SGD) algorithm. However, the SGD method doesn't consider the correlation between parameters and therefore can lead to unstable and slow convergence in training. Second-order optimization methods provide a possible solution to this issue. However these methods are either computationally heavy or do not have competitive performance. In this paper, a novel optimization method - stochastic natural gradient based on minimum variance assumption (SNGM) is proposed for training RNNLMs. It allows the natural gradient method to operate at a comparable training efficiency to the SGD method. By modifying the gradient according to the local curvature of the KL-divergence between current and updated probabilistic distributions, the proposed SNGM approach is shown to outperform both the SGD and limited memory BFGS methods across three tasks: Penn Treebank, Switchboard conversational speech recognition and AMI meeting room transcription in terms of both perplexity and word error rate." @default.
• W2940446723 created "2019-04-25" @default.
• W2940446723 creator A5004643540 @default.
• W2940446723 creator A5007473150 @default.
• W2940446723 creator A5011642893 @default.
• W2940446723 creator A5019458385 @default.
• W2940446723 creator A5025119324 @default.
• W2940446723 creator A5027263026 @default.
• W2940446723 creator A5037109470 @default.
• W2940446723 creator A5045355404 @default.
• W2940446723 date "2019-05-01" @default.
• W2940446723 modified "2023-10-14" @default.
• W2940446723 title "Recurrent Neural Network Language Model Training Using Natural Gradient" @default.
• W2940446723 cites W101286142 @default.
• W2940446723 cites W1498436455 @default.
• W2940446723 cites W1591801644 @default.
• W2940446723 cites W179875071 @default.
• W2940446723 cites W196761320 @default.
• W2940446723 cites W1970689298 @default.
• W2940446723 cites W1970789124 @default.
• W2940446723 cites W1985258458 @default.
• W2940446723 cites W2000359198 @default.
• W2940446723 cites W2006903949 @default.
• W2940446723 cites W2051434435 @default.
• W2940446723 cites W2064675550 @default.
• W2940446723 cites W2076094076 @default.
• W2940446723 cites W2077658674 @default.
• W2940446723 cites W2100506586 @default.
• W2940446723 cites W2125336414 @default.
• W2940446723 cites W2132339004 @default.
• W2940446723 cites W2167729035 @default.
• W2940446723 cites W2401430117 @default.
• W2940446723 cites W2402268235 @default.
• W2940446723 cites W2403195671 @default.
• W2940446723 cites W2537374808 @default.
• W2940446723 cites W2801666531 @default.
• W2940446723 cites W2886512927 @default.
• W2940446723 cites W2899771611 @default.
• W2940446723 cites W2963855604 @default.
• W2940446723 cites W2963857170 @default.
• W2940446723 cites W2964309400 @default.
• W2940446723 cites W3028642772 @default.
• W2940446723 cites W3086499488 @default.
• W2940446723 cites W71081281 @default.
• W2940446723 doi "https://doi.org/10.1109/icassp.2019.8682445" @default.
• W2940446723 hasPublicationYear "2019" @default.
• W2940446723 type Work @default.
• W2940446723 sameAs 2940446723 @default.
• W2940446723 citedByCount "1" @default.
• W2940446723 countsByYear W29404467232020 @default.
• W2940446723 crossrefType "proceedings-article" @default.
• W2940446723 hasAuthorship W2940446723A5004643540 @default.
• W2940446723 hasAuthorship W2940446723A5007473150 @default.
• W2940446723 hasAuthorship W2940446723A5011642893 @default.
• W2940446723 hasAuthorship W2940446723A5019458385 @default.
• W2940446723 hasAuthorship W2940446723A5025119324 @default.
• W2940446723 hasAuthorship W2940446723A5027263026 @default.
• W2940446723 hasAuthorship W2940446723A5037109470 @default.
• W2940446723 hasAuthorship W2940446723A5045355404 @default.
• W2940446723 hasConcept C100279451 @default.
• W2940446723 hasConcept C11413529 @default.
• W2940446723 hasConcept C115680565 @default.
• W2940446723 hasConcept C119857082 @default.
• W2940446723 hasConcept C137293760 @default.
• W2940446723 hasConcept C153258448 @default.
• W2940446723 hasConcept C154945302 @default.
• W2940446723 hasConcept C186644900 @default.
• W2940446723 hasConcept C206134035 @default.
• W2940446723 hasConcept C206688291 @default.
• W2940446723 hasConcept C28490314 @default.
• W2940446723 hasConcept C40969351 @default.
• W2940446723 hasConcept C41008148 @default.
• W2940446723 hasConcept C50644808 @default.
• W2940446723 hasConceptScore W2940446723C100279451 @default.
• W2940446723 hasConceptScore W2940446723C11413529 @default.
• W2940446723 hasConceptScore W2940446723C115680565 @default.
• W2940446723 hasConceptScore W2940446723C119857082 @default.
• W2940446723 hasConceptScore W2940446723C137293760 @default.
• W2940446723 hasConceptScore W2940446723C153258448 @default.
• W2940446723 hasConceptScore W2940446723C154945302 @default.
• W2940446723 hasConceptScore W2940446723C186644900 @default.
• W2940446723 hasConceptScore W2940446723C206134035 @default.
• W2940446723 hasConceptScore W2940446723C206688291 @default.
• W2940446723 hasConceptScore W2940446723C28490314 @default.
• W2940446723 hasConceptScore W2940446723C40969351 @default.
• W2940446723 hasConceptScore W2940446723C41008148 @default.
• W2940446723 hasConceptScore W2940446723C50644808 @default.
• W2940446723 hasLocation W29404467231 @default.
• W2940446723 hasOpenAccess W2940446723 @default.
• W2940446723 hasPrimaryLocation W29404467231 @default.
• W2940446723 hasRelatedWork W1989705153 @default.
• W2940446723 hasRelatedWork W2020382207 @default.
• W2940446723 hasRelatedWork W2072350286 @default.
• W2940446723 hasRelatedWork W2107734859 @default.
• W2940446723 hasRelatedWork W2159748844 @default.
• W2940446723 hasRelatedWork W2251150025 @default.
• W2940446723 hasRelatedWork W2598933668 @default.
• W2940446723 hasRelatedWork W2913200330 @default.
• W2940446723 hasRelatedWork W2940446723 @default.
• W2940446723 hasRelatedWork W4292749535 @default.

• next