Rotational Unit of Memory: A Novel Representation Unit for RNNs with Scalable Applications

<p dir="ltr">Stacking long short-term memory (LSTM) cells or gated recurrent units (GRUs) as part of a recurrent neural network (RNN) has become a standard approach to solving a number of tasks ranging from language modeling to text summarization. Although LSTMs and GRUs were designe...

وصف كامل

محفوظ في:
التفاصيل البيبلوغرافية
المؤلف الرئيسي: Rumen Dangovski (6658229) (author)
مؤلفون آخرون: Li Jing (445177) (author), Preslav Nakov (17760905) (author), Mićo Tatalović (18619243) (author), Marin Soljačić (3961958) (author)
منشور في: 2019
الموضوعات:
الوسوم: إضافة وسم
لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!
_version_ 1864513513587736576
author Rumen Dangovski (6658229)
author2 Li Jing (445177)
Preslav Nakov (17760905)
Mićo Tatalović (18619243)
Marin Soljačić (3961958)
author2_role author
author
author
author
author_facet Rumen Dangovski (6658229)
Li Jing (445177)
Preslav Nakov (17760905)
Mićo Tatalović (18619243)
Marin Soljačić (3961958)
author_role author
dc.creator.none.fl_str_mv Rumen Dangovski (6658229)
Li Jing (445177)
Preslav Nakov (17760905)
Mićo Tatalović (18619243)
Marin Soljačić (3961958)
dc.date.none.fl_str_mv 2019-04-01T03:00:00Z
dc.identifier.none.fl_str_mv 10.1162/tacl_a_00258
dc.relation.none.fl_str_mv https://figshare.com/articles/journal_contribution/Rotational_Unit_of_Memory_A_Novel_Representation_Unit_for_RNNs_with_Scalable_Applications/25908373
dc.rights.none.fl_str_mv CC BY 4.0
info:eu-repo/semantics/openAccess
dc.subject.none.fl_str_mv Information and computing sciences
Artificial intelligence
Machine learning
Language modeling
Text summarization
Long-range dependencies
Memory representation
Rotational Unit of Memory (RUM)
Unitary learning
Associative memory
dc.title.none.fl_str_mv Rotational Unit of Memory: A Novel Representation Unit for RNNs with Scalable Applications
dc.type.none.fl_str_mv Text
Journal contribution
info:eu-repo/semantics/publishedVersion
text
contribution to journal
description <p dir="ltr">Stacking long short-term memory (LSTM) cells or gated recurrent units (GRUs) as part of a recurrent neural network (RNN) has become a standard approach to solving a number of tasks ranging from language modeling to text summarization. Although LSTMs and GRUs were designed to model long-range dependencies more accurately than conventional RNNs, they nevertheless have problems copying or recalling information from the long distant past. Here, we derive a phase-coded representation of the memory state, Rotational Unit of Memory (RUM), that unifies the concepts of unitary learning and associative memory. We show experimentally that RNNs based on RUMs can solve basic sequential tasks such as memory copying and memory recall much better than LSTMs/GRUs. We further demonstrate that by replacing LSTM/GRU with RUM units we can apply neural networks to real-world problems such as language modeling and text summarization, yielding results comparable to the state of the art.</p><p><br></p><h2>Other Information</h2><p dir="ltr">Published in: Transactions of the Association for Computational Linguistics<br>License: <a href="https://creativecommons.org/licenses/by/4.0/" target="_blank">https://creativecommons.org/licenses/by/4.0/</a><br>See article on publisher's website: <a href="https://dx.doi.org/10.1162/tacl_a_00258" target="_blank">https://dx.doi.org/10.1162/tacl_a_00258</a></p>
eu_rights_str_mv openAccess
id Manara2_e99ca4ab715686fe31a81cb13e486301
identifier_str_mv 10.1162/tacl_a_00258
network_acronym_str Manara2
network_name_str Manara2
oai_identifier_str oai:figshare.com:article/25908373
publishDate 2019
repository.mail.fl_str_mv
repository.name.fl_str_mv
repository_id_str
rights_invalid_str_mv CC BY 4.0
spelling Rotational Unit of Memory: A Novel Representation Unit for RNNs with Scalable ApplicationsRumen Dangovski (6658229)Li Jing (445177)Preslav Nakov (17760905)Mićo Tatalović (18619243)Marin Soljačić (3961958)Information and computing sciencesArtificial intelligenceMachine learningLanguage modelingText summarizationLong-range dependenciesMemory representationRotational Unit of Memory (RUM)Unitary learningAssociative memory<p dir="ltr">Stacking long short-term memory (LSTM) cells or gated recurrent units (GRUs) as part of a recurrent neural network (RNN) has become a standard approach to solving a number of tasks ranging from language modeling to text summarization. Although LSTMs and GRUs were designed to model long-range dependencies more accurately than conventional RNNs, they nevertheless have problems copying or recalling information from the long distant past. Here, we derive a phase-coded representation of the memory state, Rotational Unit of Memory (RUM), that unifies the concepts of unitary learning and associative memory. We show experimentally that RNNs based on RUMs can solve basic sequential tasks such as memory copying and memory recall much better than LSTMs/GRUs. We further demonstrate that by replacing LSTM/GRU with RUM units we can apply neural networks to real-world problems such as language modeling and text summarization, yielding results comparable to the state of the art.</p><p><br></p><h2>Other Information</h2><p dir="ltr">Published in: Transactions of the Association for Computational Linguistics<br>License: <a href="https://creativecommons.org/licenses/by/4.0/" target="_blank">https://creativecommons.org/licenses/by/4.0/</a><br>See article on publisher's website: <a href="https://dx.doi.org/10.1162/tacl_a_00258" target="_blank">https://dx.doi.org/10.1162/tacl_a_00258</a></p>2019-04-01T03:00:00ZTextJournal contributioninfo:eu-repo/semantics/publishedVersiontextcontribution to journal10.1162/tacl_a_00258https://figshare.com/articles/journal_contribution/Rotational_Unit_of_Memory_A_Novel_Representation_Unit_for_RNNs_with_Scalable_Applications/25908373CC BY 4.0info:eu-repo/semantics/openAccessoai:figshare.com:article/259083732019-04-01T03:00:00Z
spellingShingle Rotational Unit of Memory: A Novel Representation Unit for RNNs with Scalable Applications
Rumen Dangovski (6658229)
Information and computing sciences
Artificial intelligence
Machine learning
Language modeling
Text summarization
Long-range dependencies
Memory representation
Rotational Unit of Memory (RUM)
Unitary learning
Associative memory
status_str publishedVersion
title Rotational Unit of Memory: A Novel Representation Unit for RNNs with Scalable Applications
title_full Rotational Unit of Memory: A Novel Representation Unit for RNNs with Scalable Applications
title_fullStr Rotational Unit of Memory: A Novel Representation Unit for RNNs with Scalable Applications
title_full_unstemmed Rotational Unit of Memory: A Novel Representation Unit for RNNs with Scalable Applications
title_short Rotational Unit of Memory: A Novel Representation Unit for RNNs with Scalable Applications
title_sort Rotational Unit of Memory: A Novel Representation Unit for RNNs with Scalable Applications
topic Information and computing sciences
Artificial intelligence
Machine learning
Language modeling
Text summarization
Long-range dependencies
Memory representation
Rotational Unit of Memory (RUM)
Unitary learning
Associative memory