MICROTRAINING FOR ITERATIVE FEW-SHOT REFINEMENT OF A NEURAL NETWORK

The disclosed microtraining techniques improve accuracy of trained neural networks by performing iterative refinement at low learning rates using a relatively short series microtraining steps. A neural network training framework receives the trained neural network along with a second training datase...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Lefohn, Aaron Eliot, Xu, Yinghao, Edelsten, Andrew Leighton, Patney, Anjul, Rowlett, Brandon Lee
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 Lefohn, Aaron Eliot
Xu, Yinghao
Edelsten, Andrew Leighton
Patney, Anjul
Rowlett, Brandon Lee
description The disclosed microtraining techniques improve accuracy of trained neural networks by performing iterative refinement at low learning rates using a relatively short series microtraining steps. A neural network training framework receives the trained neural network along with a second training dataset and set of hyperparameters. The neural network training framework produces a microtrained neural network by adjusting one or more weights of the trained neural network using a lower learning rate to facilitate incremental accuracy improvements without substantially altering the computational structure of the trained neural network. The microtrained neural network may be assessed for changes in accuracy and/or quality. Additional microtraining sessions may be performed on the microtrained neural network to further improve accuracy or quality.
format Patent
fullrecord <record><control><sourceid>epo_EVB</sourceid><recordid>TN_cdi_epo_espacenet_US2021287096A1</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>US2021287096A1</sourcerecordid><originalsourceid>FETCH-epo_espacenet_US2021287096A13</originalsourceid><addsrcrecordid>eNrjZHD29XQO8g8JcvT08_RzV3DzD1LwDHENcgzxDHNVcHMN1w328A9RCHJ18_Rz9XX1C1Hwd1NwVPBzDQ1y9AFSIeH-Qd48DKxpiTnFqbxQmptB2c01xNlDN7UgPz61uCAxOTUvtSQ-NNjIwMjQyMLcwNLM0dCYOFUA7o8skw</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>patent</recordtype></control><display><type>patent</type><title>MICROTRAINING FOR ITERATIVE FEW-SHOT REFINEMENT OF A NEURAL NETWORK</title><source>esp@cenet</source><creator>Lefohn, Aaron Eliot ; Xu, Yinghao ; Edelsten, Andrew Leighton ; Patney, Anjul ; Rowlett, Brandon Lee</creator><creatorcontrib>Lefohn, Aaron Eliot ; Xu, Yinghao ; Edelsten, Andrew Leighton ; Patney, Anjul ; Rowlett, Brandon Lee</creatorcontrib><description>The disclosed microtraining techniques improve accuracy of trained neural networks by performing iterative refinement at low learning rates using a relatively short series microtraining steps. A neural network training framework receives the trained neural network along with a second training dataset and set of hyperparameters. The neural network training framework produces a microtrained neural network by adjusting one or more weights of the trained neural network using a lower learning rate to facilitate incremental accuracy improvements without substantially altering the computational structure of the trained neural network. The microtrained neural network may be assessed for changes in accuracy and/or quality. Additional microtraining sessions may be performed on the microtrained neural network to further improve accuracy or quality.</description><language>eng</language><subject>CALCULATING ; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS ; COMPUTING ; COUNTING ; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL ; PHYSICS</subject><creationdate>2021</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&amp;date=20210916&amp;DB=EPODOC&amp;CC=US&amp;NR=2021287096A1$$EHTML$$P50$$Gepo$$Hfree_for_read</linktohtml><link.rule.ids>230,308,776,881,25542,76290</link.rule.ids><linktorsrc>$$Uhttps://worldwide.espacenet.com/publicationDetails/biblio?FT=D&amp;date=20210916&amp;DB=EPODOC&amp;CC=US&amp;NR=2021287096A1$$EView_record_in_European_Patent_Office$$FView_record_in_$$GEuropean_Patent_Office$$Hfree_for_read</linktorsrc></links><search><creatorcontrib>Lefohn, Aaron Eliot</creatorcontrib><creatorcontrib>Xu, Yinghao</creatorcontrib><creatorcontrib>Edelsten, Andrew Leighton</creatorcontrib><creatorcontrib>Patney, Anjul</creatorcontrib><creatorcontrib>Rowlett, Brandon Lee</creatorcontrib><title>MICROTRAINING FOR ITERATIVE FEW-SHOT REFINEMENT OF A NEURAL NETWORK</title><description>The disclosed microtraining techniques improve accuracy of trained neural networks by performing iterative refinement at low learning rates using a relatively short series microtraining steps. A neural network training framework receives the trained neural network along with a second training dataset and set of hyperparameters. The neural network training framework produces a microtrained neural network by adjusting one or more weights of the trained neural network using a lower learning rate to facilitate incremental accuracy improvements without substantially altering the computational structure of the trained neural network. The microtrained neural network may be assessed for changes in accuracy and/or quality. Additional microtraining sessions may be performed on the microtrained neural network to further improve accuracy or quality.</description><subject>CALCULATING</subject><subject>COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS</subject><subject>COMPUTING</subject><subject>COUNTING</subject><subject>IMAGE DATA PROCESSING OR GENERATION, IN GENERAL</subject><subject>PHYSICS</subject><fulltext>true</fulltext><rsrctype>patent</rsrctype><creationdate>2021</creationdate><recordtype>patent</recordtype><sourceid>EVB</sourceid><recordid>eNrjZHD29XQO8g8JcvT08_RzV3DzD1LwDHENcgzxDHNVcHMN1w328A9RCHJ18_Rz9XX1C1Hwd1NwVPBzDQ1y9AFSIeH-Qd48DKxpiTnFqbxQmptB2c01xNlDN7UgPz61uCAxOTUvtSQ-NNjIwMjQyMLcwNLM0dCYOFUA7o8skw</recordid><startdate>20210916</startdate><enddate>20210916</enddate><creator>Lefohn, Aaron Eliot</creator><creator>Xu, Yinghao</creator><creator>Edelsten, Andrew Leighton</creator><creator>Patney, Anjul</creator><creator>Rowlett, Brandon Lee</creator><scope>EVB</scope></search><sort><creationdate>20210916</creationdate><title>MICROTRAINING FOR ITERATIVE FEW-SHOT REFINEMENT OF A NEURAL NETWORK</title><author>Lefohn, Aaron Eliot ; Xu, Yinghao ; Edelsten, Andrew Leighton ; Patney, Anjul ; Rowlett, Brandon Lee</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-epo_espacenet_US2021287096A13</frbrgroupid><rsrctype>patents</rsrctype><prefilter>patents</prefilter><language>eng</language><creationdate>2021</creationdate><topic>CALCULATING</topic><topic>COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS</topic><topic>COMPUTING</topic><topic>COUNTING</topic><topic>IMAGE DATA PROCESSING OR GENERATION, IN GENERAL</topic><topic>PHYSICS</topic><toplevel>online_resources</toplevel><creatorcontrib>Lefohn, Aaron Eliot</creatorcontrib><creatorcontrib>Xu, Yinghao</creatorcontrib><creatorcontrib>Edelsten, Andrew Leighton</creatorcontrib><creatorcontrib>Patney, Anjul</creatorcontrib><creatorcontrib>Rowlett, Brandon Lee</creatorcontrib><collection>esp@cenet</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Lefohn, Aaron Eliot</au><au>Xu, Yinghao</au><au>Edelsten, Andrew Leighton</au><au>Patney, Anjul</au><au>Rowlett, Brandon Lee</au><format>patent</format><genre>patent</genre><ristype>GEN</ristype><title>MICROTRAINING FOR ITERATIVE FEW-SHOT REFINEMENT OF A NEURAL NETWORK</title><date>2021-09-16</date><risdate>2021</risdate><abstract>The disclosed microtraining techniques improve accuracy of trained neural networks by performing iterative refinement at low learning rates using a relatively short series microtraining steps. A neural network training framework receives the trained neural network along with a second training dataset and set of hyperparameters. The neural network training framework produces a microtrained neural network by adjusting one or more weights of the trained neural network using a lower learning rate to facilitate incremental accuracy improvements without substantially altering the computational structure of the trained neural network. The microtrained neural network may be assessed for changes in accuracy and/or quality. Additional microtraining sessions may be performed on the microtrained neural network to further improve accuracy or quality.</abstract><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier
ispartof
issn
language eng
recordid cdi_epo_espacenet_US2021287096A1
source esp@cenet
subjects CALCULATING
COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
COMPUTING
COUNTING
IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
PHYSICS
title MICROTRAINING FOR ITERATIVE FEW-SHOT REFINEMENT OF A NEURAL NETWORK
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-30T00%3A55%3A58IST&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=Lefohn,%20Aaron%20Eliot&rft.date=2021-09-16&rft_id=info:doi/&rft_dat=%3Cepo_EVB%3EUS2021287096A1%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