Neural Random-Access Machines

Neural Random-Access Machines

In this paper, we propose and investigate a new neural network architecture called Neural Random Access Machine. It can manipulate and dereference pointers to an external variable-size random-access memory. The model is trained from pure input-output examples using backpropagation. We evaluate the n...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Kurach, Karol, Andrychowicz, Marcin, Sutskever, Ilya
Format: Artikel
Sprache:eng
Schlagworte:
Computer Science - Learning
Computer Science - Neural and Evolutionary Computing
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Kurach, Karol
Andrychowicz, Marcin
Sutskever, Ilya
description In this paper, we propose and investigate a new neural network architecture called Neural Random Access Machine. It can manipulate and dereference pointers to an external variable-size random-access memory. The model is trained from pure input-output examples using backpropagation. We evaluate the new model on a number of simple algorithmic tasks whose solutions require pointer manipulation and dereferencing. Our results show that the proposed model can learn to solve algorithmic tasks of such type and is capable of operating on simple data structures like linked-lists and binary trees. For easier tasks, the learned solutions generalize to sequences of arbitrary length. Moreover, memory access during inference can be done in a constant time under some assumptions.
doi_str_mv 10.48550/arxiv.1511.06392
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_1511_06392</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>1511_06392</sourcerecordid><originalsourceid>FETCH-LOGICAL-a672-4165617571935f07cf5ae5b767779e77d071ddff51523fd2c0f9cafafa440f983</originalsourceid><addsrcrecordid>eNotzskKwkAQBNC5eJDoB3gQ8wOJ05n0tDmKuIELSO6hnQUDbiQo-vfGKHWoOhVPiAHIOJ0gyjFXr_IZAwLEUqss6Yrhzj0qPocHvtrbJZoa4-o63LI5lVdX90TH87l2_X8HIl_M89kq2uyX69l0E7GmJEpBowZCgkyhl2Q8ssMjaSLKHJGVBNZ6j4CJ8jYx0meGfZM0beZEBWL0u219xb0qL1y9i6-zaJ3qA2wXNjg</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Neural Random-Access Machines</title><source>arXiv.org</source><creator>Kurach, Karol ; Andrychowicz, Marcin ; Sutskever, Ilya</creator><creatorcontrib>Kurach, Karol ; Andrychowicz, Marcin ; Sutskever, Ilya</creatorcontrib><description>In this paper, we propose and investigate a new neural network architecture called Neural Random Access Machine. It can manipulate and dereference pointers to an external variable-size random-access memory. The model is trained from pure input-output examples using backpropagation. We evaluate the new model on a number of simple algorithmic tasks whose solutions require pointer manipulation and dereferencing. Our results show that the proposed model can learn to solve algorithmic tasks of such type and is capable of operating on simple data structures like linked-lists and binary trees. For easier tasks, the learned solutions generalize to sequences of arbitrary length. Moreover, memory access during inference can be done in a constant time under some assumptions.</description><identifier>DOI: 10.48550/arxiv.1511.06392</identifier><language>eng</language><subject>Computer Science - Learning ; Computer Science - Neural and Evolutionary Computing</subject><creationdate>2015-11</creationdate><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,776,881</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/1511.06392$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.1511.06392$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Kurach, Karol</creatorcontrib><creatorcontrib>Andrychowicz, Marcin</creatorcontrib><creatorcontrib>Sutskever, Ilya</creatorcontrib><title>Neural Random-Access Machines</title><description>In this paper, we propose and investigate a new neural network architecture called Neural Random Access Machine. It can manipulate and dereference pointers to an external variable-size random-access memory. The model is trained from pure input-output examples using backpropagation. We evaluate the new model on a number of simple algorithmic tasks whose solutions require pointer manipulation and dereferencing. Our results show that the proposed model can learn to solve algorithmic tasks of such type and is capable of operating on simple data structures like linked-lists and binary trees. For easier tasks, the learned solutions generalize to sequences of arbitrary length. Moreover, memory access during inference can be done in a constant time under some assumptions.</description><subject>Computer Science - Learning</subject><subject>Computer Science - Neural and Evolutionary Computing</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2015</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotzskKwkAQBNC5eJDoB3gQ8wOJ05n0tDmKuIELSO6hnQUDbiQo-vfGKHWoOhVPiAHIOJ0gyjFXr_IZAwLEUqss6Yrhzj0qPocHvtrbJZoa4-o63LI5lVdX90TH87l2_X8HIl_M89kq2uyX69l0E7GmJEpBowZCgkyhl2Q8ssMjaSLKHJGVBNZ6j4CJ8jYx0meGfZM0beZEBWL0u219xb0qL1y9i6-zaJ3qA2wXNjg</recordid><startdate>20151119</startdate><enddate>20151119</enddate><creator>Kurach, Karol</creator><creator>Andrychowicz, Marcin</creator><creator>Sutskever, Ilya</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20151119</creationdate><title>Neural Random-Access Machines</title><author>Kurach, Karol ; Andrychowicz, Marcin ; Sutskever, Ilya</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a672-4165617571935f07cf5ae5b767779e77d071ddff51523fd2c0f9cafafa440f983</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2015</creationdate><topic>Computer Science - Learning</topic><topic>Computer Science - Neural and Evolutionary Computing</topic><toplevel>online_resources</toplevel><creatorcontrib>Kurach, Karol</creatorcontrib><creatorcontrib>Andrychowicz, Marcin</creatorcontrib><creatorcontrib>Sutskever, Ilya</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Kurach, Karol</au><au>Andrychowicz, Marcin</au><au>Sutskever, Ilya</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Neural Random-Access Machines</atitle><date>2015-11-19</date><risdate>2015</risdate><abstract>In this paper, we propose and investigate a new neural network architecture called Neural Random Access Machine. It can manipulate and dereference pointers to an external variable-size random-access memory. The model is trained from pure input-output examples using backpropagation. We evaluate the new model on a number of simple algorithmic tasks whose solutions require pointer manipulation and dereferencing. Our results show that the proposed model can learn to solve algorithmic tasks of such type and is capable of operating on simple data structures like linked-lists and binary trees. For easier tasks, the learned solutions generalize to sequences of arbitrary length. Moreover, memory access during inference can be done in a constant time under some assumptions.</abstract><doi>10.48550/arxiv.1511.06392</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.1511.06392
ispartof
issn
language eng
recordid cdi_arxiv_primary_1511_06392
source arXiv.org
subjects Computer Science - Learning
Computer Science - Neural and Evolutionary Computing
title Neural Random-Access Machines
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-23T16%3A10%3A24IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Neural%20Random-Access%20Machines&rft.au=Kurach,%20Karol&rft.date=2015-11-19&rft_id=info:doi/10.48550/arxiv.1511.06392&rft_dat=%3Carxiv_GOX%3E1511_06392%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true