Transcription of Sequence to Sequence Learning with Neural …
{{id}} {{{paragraph}}}
Example: barber
Sequence to Sequence Learning with Neural …
[ ] 14 Dec 2014 Sequence to Sequence Learningwith Neural NetworksIlya V. Neural Networks (DNNs) are powerful models that have achieved excel-lent performance on difficult Learning tasks. Although DNNswork well wheneverlarge labeled training sets are available, they cannot be used to map sequences tosequences. In this paper, we present a general end-to-end approach to sequencelearning that makes minimal assumptions on the Sequence structure. Our methoduses a multilayered Long Short-Term Memory (LSTM) to map theinput sequenceto a vector of a fixed dimensionality, and then another deep LSTM to decode thetarget Sequence from the vector. Our main result is that on anEnglish to Frenchtranslation task from the WMT 14 dataset, the translationsproduced by the LSTM achieve a BLEU score of on the entire test set, where the LSTM s BLEU score was penalized on out-of-vocabulary words.
arXiv:1409.3215v3 [cs.CL] 14 Dec 2014 Sequence to Sequence Learning with Neural Networks Ilya Sutskever Google ilyasu@google.com Oriol Vinyals Google
Loading..
Tags:
Information
Domain:
Source:
Link to this page:
Please notify us if you found a problem with this document:
Related search queries
Binarized Neural Networks: Training Neural, Binarized Neural Networks: Training Neural Networks, Artificial Neural Networks, Neural Networks, MatConvNet: Convolutional Neural Networks for, Neural Networks for High Dimension, Low, Forecasting with Artificial Neural Networks, Neural, GPUs in Accelerating Next, GPUs in Accelerating Next-Generation Deep Neural Networks, Neural Networks: MATLAB examples, Neural Networks: MATLAB examples Neural Networks, Facial Expression Recognizer using, Facial Expression Recognizer using Deep Neural, AT74 ARTIFICIAL INTELLIGENCE & NEURAL