Trainable Communication Systems: Concepts and Prototype
We consider a trainable point-to-point communication system, where both transmitter and receiver are implemented as neural networks (NNs), and demonstrate that training on the bit-wise mutual information (BMI) allows seamless integration with practical bit-metric decoding (BMD) receivers, as well as...
Gespeichert in:
| Veröffentlicht in: | IEEE transactions on communications 2020-09, Vol.68 (9), p.5489-5503 |
|---|---|
| Hauptverfasser: | , , , , , |
| Format: | Artikel |
| Sprache: | eng |
| Schlagworte: | |
| Online-Zugang: | Volltext bestellen |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| container_end_page | 5503 |
|---|---|
| container_issue | 9 |
| container_start_page | 5489 |
| container_title | IEEE transactions on communications |
| container_volume | 68 |
| creator | Cammerer, Sebastian Aoudia, Faycal Ait Dorner, Sebastian Stark, Maximilian Hoydis, Jakob ten Brink, Stephan |
| description | We consider a trainable point-to-point communication system, where both transmitter and receiver are implemented as neural networks (NNs), and demonstrate that training on the bit-wise mutual information (BMI) allows seamless integration with practical bit-metric decoding (BMD) receivers, as well as joint optimization of constellation shaping and labeling. Moreover, we present a fully differentiable neural iterative demapping and decoding (IDD) structure which achieves significant gains on additive white Gaussian noise (AWGN) channels using a standard 802.11n low-density parity-check (LDPC) code. The strength of this approach is that it can be applied to arbitrary channels without any modifications. Going one step further, we show that careful code design can lead to further performance improvements. Lastly, we show the viability of the proposed system through implementation on software-defined radios (SDRs) and training of the end-to-end system on the actual wireless channel. Experimental results reveal that the proposed method enables significant gains compared to conventional techniques. |
| doi_str_mv | 10.1109/TCOMM.2020.3002915 |
| format | Article |
| fullrecord | <record><control><sourceid>proquest_RIE</sourceid><recordid>TN_cdi_proquest_journals_2444610139</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><ieee_id>9118963</ieee_id><sourcerecordid>2444610139</sourcerecordid><originalsourceid>FETCH-LOGICAL-c344t-cef4a93d7dd0f212bf7d393ee2c166ae6e2580d4c60107b309670182a2d4083e3</originalsourceid><addsrcrecordid>eNo9kEtLAzEUhYMoWKt_QDcDrqfePCYPdzJYFVoqWNchzdyBKZ2HSbrov3dqi6sLh_udAx8h9xRmlIJ5Wper5XLGgMGMAzBDiwsyoUWhc9CFuiQTAAO5VEpfk5sYtwAggPMJUevgms5tdpiVfdvuu8a71PRd9nWICdv4PMadxyHFzHVV9hn61KfDgLfkqna7iHfnOyXf89d1-Z4vVm8f5csi91yIlHushTO8UlUFNaNsU6uKG47IPJXSoURWaKiEl0BBbTgYqYBq5lglQHPkU_J46h1C_7PHmOy234dunLRMCCEp0LFvStjpy4c-xoC1HULTunCwFOxRkP0TZI-C7FnQCD2coAYR_wFDqTaS819bTWC7</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2444610139</pqid></control><display><type>article</type><title>Trainable Communication Systems: Concepts and Prototype</title><source>IEEE Electronic Library (IEL)</source><creator>Cammerer, Sebastian ; Aoudia, Faycal Ait ; Dorner, Sebastian ; Stark, Maximilian ; Hoydis, Jakob ; ten Brink, Stephan</creator><creatorcontrib>Cammerer, Sebastian ; Aoudia, Faycal Ait ; Dorner, Sebastian ; Stark, Maximilian ; Hoydis, Jakob ; ten Brink, Stephan</creatorcontrib><description>We consider a trainable point-to-point communication system, where both transmitter and receiver are implemented as neural networks (NNs), and demonstrate that training on the bit-wise mutual information (BMI) allows seamless integration with practical bit-metric decoding (BMD) receivers, as well as joint optimization of constellation shaping and labeling. Moreover, we present a fully differentiable neural iterative demapping and decoding (IDD) structure which achieves significant gains on additive white Gaussian noise (AWGN) channels using a standard 802.11n low-density parity-check (LDPC) code. The strength of this approach is that it can be applied to arbitrary channels without any modifications. Going one step further, we show that careful code design can lead to further performance improvements. Lastly, we show the viability of the proposed system through implementation on software-defined radios (SDRs) and training of the end-to-end system on the actual wireless channel. Experimental results reveal that the proposed method enables significant gains compared to conventional techniques.</description><identifier>ISSN: 0090-6778</identifier><identifier>EISSN: 1558-0857</identifier><identifier>DOI: 10.1109/TCOMM.2020.3002915</identifier><identifier>CODEN: IECMBT</identifier><language>eng</language><publisher>New York: IEEE</publisher><subject>Autoencoder ; Channels ; code design ; Codes ; Communication systems ; Communications systems ; Constellations ; Decoding ; Design modifications ; end-to-end learning ; Error correcting codes ; geometric shaping ; Iterative decoding ; iterative demapping and decoding ; Neural networks ; Optical transmitters ; Optimization ; Random noise ; Receivers ; Software radio ; software-defined radio ; Training</subject><ispartof>IEEE transactions on communications, 2020-09, Vol.68 (9), p.5489-5503</ispartof><rights>Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2020</rights><lds50>peer_reviewed</lds50><woscitedreferencessubscribed>false</woscitedreferencessubscribed><citedby>FETCH-LOGICAL-c344t-cef4a93d7dd0f212bf7d393ee2c166ae6e2580d4c60107b309670182a2d4083e3</citedby><cites>FETCH-LOGICAL-c344t-cef4a93d7dd0f212bf7d393ee2c166ae6e2580d4c60107b309670182a2d4083e3</cites><orcidid>0000-0003-2723-6171 ; 0000-0002-0438-967X ; 0000-0002-1750-5895 ; 0000-0002-7614-5156 ; 0000-0003-4396-1082 ; 0000-0003-1502-2571</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://ieeexplore.ieee.org/document/9118963$$EHTML$$P50$$Gieee$$H</linktohtml><link.rule.ids>314,776,780,792,27901,27902,54733</link.rule.ids><linktorsrc>$$Uhttps://ieeexplore.ieee.org/document/9118963$$EView_record_in_IEEE$$FView_record_in_$$GIEEE</linktorsrc></links><search><creatorcontrib>Cammerer, Sebastian</creatorcontrib><creatorcontrib>Aoudia, Faycal Ait</creatorcontrib><creatorcontrib>Dorner, Sebastian</creatorcontrib><creatorcontrib>Stark, Maximilian</creatorcontrib><creatorcontrib>Hoydis, Jakob</creatorcontrib><creatorcontrib>ten Brink, Stephan</creatorcontrib><title>Trainable Communication Systems: Concepts and Prototype</title><title>IEEE transactions on communications</title><addtitle>TCOMM</addtitle><description>We consider a trainable point-to-point communication system, where both transmitter and receiver are implemented as neural networks (NNs), and demonstrate that training on the bit-wise mutual information (BMI) allows seamless integration with practical bit-metric decoding (BMD) receivers, as well as joint optimization of constellation shaping and labeling. Moreover, we present a fully differentiable neural iterative demapping and decoding (IDD) structure which achieves significant gains on additive white Gaussian noise (AWGN) channels using a standard 802.11n low-density parity-check (LDPC) code. The strength of this approach is that it can be applied to arbitrary channels without any modifications. Going one step further, we show that careful code design can lead to further performance improvements. Lastly, we show the viability of the proposed system through implementation on software-defined radios (SDRs) and training of the end-to-end system on the actual wireless channel. Experimental results reveal that the proposed method enables significant gains compared to conventional techniques.</description><subject>Autoencoder</subject><subject>Channels</subject><subject>code design</subject><subject>Codes</subject><subject>Communication systems</subject><subject>Communications systems</subject><subject>Constellations</subject><subject>Decoding</subject><subject>Design modifications</subject><subject>end-to-end learning</subject><subject>Error correcting codes</subject><subject>geometric shaping</subject><subject>Iterative decoding</subject><subject>iterative demapping and decoding</subject><subject>Neural networks</subject><subject>Optical transmitters</subject><subject>Optimization</subject><subject>Random noise</subject><subject>Receivers</subject><subject>Software radio</subject><subject>software-defined radio</subject><subject>Training</subject><issn>0090-6778</issn><issn>1558-0857</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2020</creationdate><recordtype>article</recordtype><sourceid>RIE</sourceid><recordid>eNo9kEtLAzEUhYMoWKt_QDcDrqfePCYPdzJYFVoqWNchzdyBKZ2HSbrov3dqi6sLh_udAx8h9xRmlIJ5Wper5XLGgMGMAzBDiwsyoUWhc9CFuiQTAAO5VEpfk5sYtwAggPMJUevgms5tdpiVfdvuu8a71PRd9nWICdv4PMadxyHFzHVV9hn61KfDgLfkqna7iHfnOyXf89d1-Z4vVm8f5csi91yIlHushTO8UlUFNaNsU6uKG47IPJXSoURWaKiEl0BBbTgYqYBq5lglQHPkU_J46h1C_7PHmOy234dunLRMCCEp0LFvStjpy4c-xoC1HULTunCwFOxRkP0TZI-C7FnQCD2coAYR_wFDqTaS819bTWC7</recordid><startdate>20200901</startdate><enddate>20200901</enddate><creator>Cammerer, Sebastian</creator><creator>Aoudia, Faycal Ait</creator><creator>Dorner, Sebastian</creator><creator>Stark, Maximilian</creator><creator>Hoydis, Jakob</creator><creator>ten Brink, Stephan</creator><general>IEEE</general><general>The Institute of Electrical and Electronics Engineers, Inc. (IEEE)</general><scope>97E</scope><scope>RIA</scope><scope>RIE</scope><scope>AAYXX</scope><scope>CITATION</scope><scope>7SP</scope><scope>8FD</scope><scope>L7M</scope><orcidid>https://orcid.org/0000-0003-2723-6171</orcidid><orcidid>https://orcid.org/0000-0002-0438-967X</orcidid><orcidid>https://orcid.org/0000-0002-1750-5895</orcidid><orcidid>https://orcid.org/0000-0002-7614-5156</orcidid><orcidid>https://orcid.org/0000-0003-4396-1082</orcidid><orcidid>https://orcid.org/0000-0003-1502-2571</orcidid></search><sort><creationdate>20200901</creationdate><title>Trainable Communication Systems: Concepts and Prototype</title><author>Cammerer, Sebastian ; Aoudia, Faycal Ait ; Dorner, Sebastian ; Stark, Maximilian ; Hoydis, Jakob ; ten Brink, Stephan</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c344t-cef4a93d7dd0f212bf7d393ee2c166ae6e2580d4c60107b309670182a2d4083e3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2020</creationdate><topic>Autoencoder</topic><topic>Channels</topic><topic>code design</topic><topic>Codes</topic><topic>Communication systems</topic><topic>Communications systems</topic><topic>Constellations</topic><topic>Decoding</topic><topic>Design modifications</topic><topic>end-to-end learning</topic><topic>Error correcting codes</topic><topic>geometric shaping</topic><topic>Iterative decoding</topic><topic>iterative demapping and decoding</topic><topic>Neural networks</topic><topic>Optical transmitters</topic><topic>Optimization</topic><topic>Random noise</topic><topic>Receivers</topic><topic>Software radio</topic><topic>software-defined radio</topic><topic>Training</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Cammerer, Sebastian</creatorcontrib><creatorcontrib>Aoudia, Faycal Ait</creatorcontrib><creatorcontrib>Dorner, Sebastian</creatorcontrib><creatorcontrib>Stark, Maximilian</creatorcontrib><creatorcontrib>Hoydis, Jakob</creatorcontrib><creatorcontrib>ten Brink, Stephan</creatorcontrib><collection>IEEE All-Society Periodicals Package (ASPP) 2005-present</collection><collection>IEEE All-Society Periodicals Package (ASPP) 1998-Present</collection><collection>IEEE Electronic Library (IEL)</collection><collection>CrossRef</collection><collection>Electronics & Communications Abstracts</collection><collection>Technology Research Database</collection><collection>Advanced Technologies Database with Aerospace</collection><jtitle>IEEE transactions on communications</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Cammerer, Sebastian</au><au>Aoudia, Faycal Ait</au><au>Dorner, Sebastian</au><au>Stark, Maximilian</au><au>Hoydis, Jakob</au><au>ten Brink, Stephan</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Trainable Communication Systems: Concepts and Prototype</atitle><jtitle>IEEE transactions on communications</jtitle><stitle>TCOMM</stitle><date>2020-09-01</date><risdate>2020</risdate><volume>68</volume><issue>9</issue><spage>5489</spage><epage>5503</epage><pages>5489-5503</pages><issn>0090-6778</issn><eissn>1558-0857</eissn><coden>IECMBT</coden><abstract>We consider a trainable point-to-point communication system, where both transmitter and receiver are implemented as neural networks (NNs), and demonstrate that training on the bit-wise mutual information (BMI) allows seamless integration with practical bit-metric decoding (BMD) receivers, as well as joint optimization of constellation shaping and labeling. Moreover, we present a fully differentiable neural iterative demapping and decoding (IDD) structure which achieves significant gains on additive white Gaussian noise (AWGN) channels using a standard 802.11n low-density parity-check (LDPC) code. The strength of this approach is that it can be applied to arbitrary channels without any modifications. Going one step further, we show that careful code design can lead to further performance improvements. Lastly, we show the viability of the proposed system through implementation on software-defined radios (SDRs) and training of the end-to-end system on the actual wireless channel. Experimental results reveal that the proposed method enables significant gains compared to conventional techniques.</abstract><cop>New York</cop><pub>IEEE</pub><doi>10.1109/TCOMM.2020.3002915</doi><tpages>15</tpages><orcidid>https://orcid.org/0000-0003-2723-6171</orcidid><orcidid>https://orcid.org/0000-0002-0438-967X</orcidid><orcidid>https://orcid.org/0000-0002-1750-5895</orcidid><orcidid>https://orcid.org/0000-0002-7614-5156</orcidid><orcidid>https://orcid.org/0000-0003-4396-1082</orcidid><orcidid>https://orcid.org/0000-0003-1502-2571</orcidid></addata></record> |
| fulltext | fulltext_linktorsrc |
| identifier | ISSN: 0090-6778 |
| ispartof | IEEE transactions on communications, 2020-09, Vol.68 (9), p.5489-5503 |
| issn | 0090-6778 1558-0857 |
| language | eng |
| recordid | cdi_proquest_journals_2444610139 |
| source | IEEE Electronic Library (IEL) |
| subjects | Autoencoder Channels code design Codes Communication systems Communications systems Constellations Decoding Design modifications end-to-end learning Error correcting codes geometric shaping Iterative decoding iterative demapping and decoding Neural networks Optical transmitters Optimization Random noise Receivers Software radio software-defined radio Training |
| title | Trainable Communication Systems: Concepts and Prototype |
| url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-05T18%3A56%3A34IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_RIE&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Trainable%20Communication%20Systems:%20Concepts%20and%20Prototype&rft.jtitle=IEEE%20transactions%20on%20communications&rft.au=Cammerer,%20Sebastian&rft.date=2020-09-01&rft.volume=68&rft.issue=9&rft.spage=5489&rft.epage=5503&rft.pages=5489-5503&rft.issn=0090-6778&rft.eissn=1558-0857&rft.coden=IECMBT&rft_id=info:doi/10.1109/TCOMM.2020.3002915&rft_dat=%3Cproquest_RIE%3E2444610139%3C/proquest_RIE%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2444610139&rft_id=info:pmid/&rft_ieee_id=9118963&rfr_iscdi=true |