Controlling agents over long time scales using temporal value transport
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network system used to control an agent interacting with an environment to perform a specified task. One of the methods includes causing the agent to perform a task episode in which...
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Patent |
| Sprache: | eng |
| Schlagworte: | |
| 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 | Wayne, Gregory Duncan Abramson, Joshua Simon Hung, Chia-Chun Lillicrap, Timothy Paul |
| description | Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network system used to control an agent interacting with an environment to perform a specified task. One of the methods includes causing the agent to perform a task episode in which the agent attempts to perform the specified task; for each of one or more particular time steps in the sequence: generating a modified reward for the particular time step from (i) the actual reward at the time step and (ii) value predictions at one or more time steps that are more than a threshold number of time steps after the particular time step in the sequence; and training, through reinforcement learning, the neural network system using at least the modified rewards for the particular time steps. |
| format | Patent |
| fullrecord | <record><control><sourceid>epo_EVB</sourceid><recordid>TN_cdi_epo_espacenet_US10789511B2</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>US10789511B2</sourcerecordid><originalsourceid>FETCH-epo_espacenet_US10789511B23</originalsourceid><addsrcrecordid>eNqNyk0KwkAMhuFuXIh6h3gAwVHEurX4s1fXJZTPUkgnwyTt-a3gAVy98PDOi1ul0bOKdLElbhHdSEdkEp3Aux5kDQuMBvsujj5pZqGRZQB55mgT-LKYvVkMq18Xxfp6eVb3DZLWsMQNIrx-PcL2WJ4OIZx3-3-eD7M5NLk</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>patent</recordtype></control><display><type>patent</type><title>Controlling agents over long time scales using temporal value transport</title><source>esp@cenet</source><creator>Wayne, Gregory Duncan ; Abramson, Joshua Simon ; Hung, Chia-Chun ; Lillicrap, Timothy Paul</creator><creatorcontrib>Wayne, Gregory Duncan ; Abramson, Joshua Simon ; Hung, Chia-Chun ; Lillicrap, Timothy Paul</creatorcontrib><description>Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network system used to control an agent interacting with an environment to perform a specified task. One of the methods includes causing the agent to perform a task episode in which the agent attempts to perform the specified task; for each of one or more particular time steps in the sequence: generating a modified reward for the particular time step from (i) the actual reward at the time step and (ii) value predictions at one or more time steps that are more than a threshold number of time steps after the particular time step in the sequence; and training, through reinforcement learning, the neural network system using at least the modified rewards for the particular time steps.</description><language>eng</language><subject>CALCULATING ; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS ; COMPUTING ; COUNTING ; ELECTRIC DIGITAL DATA PROCESSING ; PHYSICS</subject><creationdate>2020</creationdate><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://worldwide.espacenet.com/publicationDetails/biblio?FT=D&date=20200929&DB=EPODOC&CC=US&NR=10789511B2$$EHTML$$P50$$Gepo$$Hfree_for_read</linktohtml><link.rule.ids>230,308,776,881,25543,76293</link.rule.ids><linktorsrc>$$Uhttps://worldwide.espacenet.com/publicationDetails/biblio?FT=D&date=20200929&DB=EPODOC&CC=US&NR=10789511B2$$EView_record_in_European_Patent_Office$$FView_record_in_$$GEuropean_Patent_Office$$Hfree_for_read</linktorsrc></links><search><creatorcontrib>Wayne, Gregory Duncan</creatorcontrib><creatorcontrib>Abramson, Joshua Simon</creatorcontrib><creatorcontrib>Hung, Chia-Chun</creatorcontrib><creatorcontrib>Lillicrap, Timothy Paul</creatorcontrib><title>Controlling agents over long time scales using temporal value transport</title><description>Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network system used to control an agent interacting with an environment to perform a specified task. One of the methods includes causing the agent to perform a task episode in which the agent attempts to perform the specified task; for each of one or more particular time steps in the sequence: generating a modified reward for the particular time step from (i) the actual reward at the time step and (ii) value predictions at one or more time steps that are more than a threshold number of time steps after the particular time step in the sequence; and training, through reinforcement learning, the neural network system using at least the modified rewards for the particular time steps.</description><subject>CALCULATING</subject><subject>COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS</subject><subject>COMPUTING</subject><subject>COUNTING</subject><subject>ELECTRIC DIGITAL DATA PROCESSING</subject><subject>PHYSICS</subject><fulltext>true</fulltext><rsrctype>patent</rsrctype><creationdate>2020</creationdate><recordtype>patent</recordtype><sourceid>EVB</sourceid><recordid>eNqNyk0KwkAMhuFuXIh6h3gAwVHEurX4s1fXJZTPUkgnwyTt-a3gAVy98PDOi1ul0bOKdLElbhHdSEdkEp3Aux5kDQuMBvsujj5pZqGRZQB55mgT-LKYvVkMq18Xxfp6eVb3DZLWsMQNIrx-PcL2WJ4OIZx3-3-eD7M5NLk</recordid><startdate>20200929</startdate><enddate>20200929</enddate><creator>Wayne, Gregory Duncan</creator><creator>Abramson, Joshua Simon</creator><creator>Hung, Chia-Chun</creator><creator>Lillicrap, Timothy Paul</creator><scope>EVB</scope></search><sort><creationdate>20200929</creationdate><title>Controlling agents over long time scales using temporal value transport</title><author>Wayne, Gregory Duncan ; Abramson, Joshua Simon ; Hung, Chia-Chun ; Lillicrap, Timothy Paul</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-epo_espacenet_US10789511B23</frbrgroupid><rsrctype>patents</rsrctype><prefilter>patents</prefilter><language>eng</language><creationdate>2020</creationdate><topic>CALCULATING</topic><topic>COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS</topic><topic>COMPUTING</topic><topic>COUNTING</topic><topic>ELECTRIC DIGITAL DATA PROCESSING</topic><topic>PHYSICS</topic><toplevel>online_resources</toplevel><creatorcontrib>Wayne, Gregory Duncan</creatorcontrib><creatorcontrib>Abramson, Joshua Simon</creatorcontrib><creatorcontrib>Hung, Chia-Chun</creatorcontrib><creatorcontrib>Lillicrap, Timothy Paul</creatorcontrib><collection>esp@cenet</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Wayne, Gregory Duncan</au><au>Abramson, Joshua Simon</au><au>Hung, Chia-Chun</au><au>Lillicrap, Timothy Paul</au><format>patent</format><genre>patent</genre><ristype>GEN</ristype><title>Controlling agents over long time scales using temporal value transport</title><date>2020-09-29</date><risdate>2020</risdate><abstract>Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network system used to control an agent interacting with an environment to perform a specified task. One of the methods includes causing the agent to perform a task episode in which the agent attempts to perform the specified task; for each of one or more particular time steps in the sequence: generating a modified reward for the particular time step from (i) the actual reward at the time step and (ii) value predictions at one or more time steps that are more than a threshold number of time steps after the particular time step in the sequence; and training, through reinforcement learning, the neural network system using at least the modified rewards for the particular time steps.</abstract><oa>free_for_read</oa></addata></record> |
| fulltext | fulltext_linktorsrc |
| identifier | |
| ispartof | |
| issn | |
| language | eng |
| recordid | cdi_epo_espacenet_US10789511B2 |
| source | esp@cenet |
| subjects | CALCULATING COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS COMPUTING COUNTING ELECTRIC DIGITAL DATA PROCESSING PHYSICS |
| title | Controlling agents over long time scales using temporal value transport |
| url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-25T04%3A05%3A18IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-epo_EVB&rft_val_fmt=info:ofi/fmt:kev:mtx:patent&rft.genre=patent&rft.au=Wayne,%20Gregory%20Duncan&rft.date=2020-09-29&rft_id=info:doi/&rft_dat=%3Cepo_EVB%3EUS10789511B2%3C/epo_EVB%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 |