Should We Be Pre-training? An Argument for End-task Aware Training as an Alternative
In most settings of practical concern, machine learning practitioners know in advance what end-task they wish to boost with auxiliary tasks. However, widely used methods for leveraging auxiliary data like pre-training and its continued-pretraining variant are end-task agnostic: they rarely, if ever,...
Gespeichert in:
Veröffentlicht in: | arXiv.org 2022-02 |
---|---|
Hauptverfasser: | , , , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
container_end_page | |
---|---|
container_issue | |
container_start_page | |
container_title | arXiv.org |
container_volume | |
creator | Dery, Lucio M Michel, Paul Talwalkar, Ameet Neubig, Graham |
description | In most settings of practical concern, machine learning practitioners know in advance what end-task they wish to boost with auxiliary tasks. However, widely used methods for leveraging auxiliary data like pre-training and its continued-pretraining variant are end-task agnostic: they rarely, if ever, exploit knowledge of the target task. We study replacing end-task agnostic continued training of pre-trained language models with end-task aware training of said models. We argue that for sufficiently important end-tasks, the benefits of leveraging auxiliary data in a task-aware fashion can justify forgoing the traditional approach of obtaining generic, end-task agnostic representations as with (continued) pre-training. On three different low-resource NLP tasks from two domains, we demonstrate that multi-tasking the end-task and auxiliary objectives results in significantly better downstream task performance than the widely-used task-agnostic continued pre-training paradigm of Gururangan et al. (2020). We next introduce an online meta-learning algorithm that learns a set of multi-task weights to better balance among our multiple auxiliary objectives, achieving further improvements on end-task performance and data efficiency. |
format | Article |
fullrecord | <record><control><sourceid>proquest</sourceid><recordid>TN_cdi_proquest_journals_2573163378</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2573163378</sourcerecordid><originalsourceid>FETCH-proquest_journals_25731633783</originalsourceid><addsrcrecordid>eNqNyk0LgjAYAOARBEn5H17oPNCtqbewMDoGCR1l4Ktpa6t91N_Pgz-g03N5FiRinKe02DG2IrFzY5IkLMuZEDwi9fVugmrhhnBAuFik3spBD7rfQ6mhtH14ovbQGQuVbqmX7gHlV1qEeo4gHcipKo9WSz98cEOWnVQO49k12Z6q-nimL2veAZ1vRhOmq1zDRM7TjPO84P-tH5huPvI</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2573163378</pqid></control><display><type>article</type><title>Should We Be Pre-training? An Argument for End-task Aware Training as an Alternative</title><source>Free E- Journals</source><creator>Dery, Lucio M ; Michel, Paul ; Talwalkar, Ameet ; Neubig, Graham</creator><creatorcontrib>Dery, Lucio M ; Michel, Paul ; Talwalkar, Ameet ; Neubig, Graham</creatorcontrib><description>In most settings of practical concern, machine learning practitioners know in advance what end-task they wish to boost with auxiliary tasks. However, widely used methods for leveraging auxiliary data like pre-training and its continued-pretraining variant are end-task agnostic: they rarely, if ever, exploit knowledge of the target task. We study replacing end-task agnostic continued training of pre-trained language models with end-task aware training of said models. We argue that for sufficiently important end-tasks, the benefits of leveraging auxiliary data in a task-aware fashion can justify forgoing the traditional approach of obtaining generic, end-task agnostic representations as with (continued) pre-training. On three different low-resource NLP tasks from two domains, we demonstrate that multi-tasking the end-task and auxiliary objectives results in significantly better downstream task performance than the widely-used task-agnostic continued pre-training paradigm of Gururangan et al. (2020). We next introduce an online meta-learning algorithm that learns a set of multi-task weights to better balance among our multiple auxiliary objectives, achieving further improvements on end-task performance and data efficiency.</description><identifier>EISSN: 2331-8422</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Algorithms ; Downstream effects ; Machine learning ; Multitasking ; Training</subject><ispartof>arXiv.org, 2022-02</ispartof><rights>2022. This work is published under http://arxiv.org/licenses/nonexclusive-distrib/1.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.</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>776,780</link.rule.ids></links><search><creatorcontrib>Dery, Lucio M</creatorcontrib><creatorcontrib>Michel, Paul</creatorcontrib><creatorcontrib>Talwalkar, Ameet</creatorcontrib><creatorcontrib>Neubig, Graham</creatorcontrib><title>Should We Be Pre-training? An Argument for End-task Aware Training as an Alternative</title><title>arXiv.org</title><description>In most settings of practical concern, machine learning practitioners know in advance what end-task they wish to boost with auxiliary tasks. However, widely used methods for leveraging auxiliary data like pre-training and its continued-pretraining variant are end-task agnostic: they rarely, if ever, exploit knowledge of the target task. We study replacing end-task agnostic continued training of pre-trained language models with end-task aware training of said models. We argue that for sufficiently important end-tasks, the benefits of leveraging auxiliary data in a task-aware fashion can justify forgoing the traditional approach of obtaining generic, end-task agnostic representations as with (continued) pre-training. On three different low-resource NLP tasks from two domains, we demonstrate that multi-tasking the end-task and auxiliary objectives results in significantly better downstream task performance than the widely-used task-agnostic continued pre-training paradigm of Gururangan et al. (2020). We next introduce an online meta-learning algorithm that learns a set of multi-task weights to better balance among our multiple auxiliary objectives, achieving further improvements on end-task performance and data efficiency.</description><subject>Algorithms</subject><subject>Downstream effects</subject><subject>Machine learning</subject><subject>Multitasking</subject><subject>Training</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2022</creationdate><recordtype>article</recordtype><sourceid>BENPR</sourceid><recordid>eNqNyk0LgjAYAOARBEn5H17oPNCtqbewMDoGCR1l4Ktpa6t91N_Pgz-g03N5FiRinKe02DG2IrFzY5IkLMuZEDwi9fVugmrhhnBAuFik3spBD7rfQ6mhtH14ovbQGQuVbqmX7gHlV1qEeo4gHcipKo9WSz98cEOWnVQO49k12Z6q-nimL2veAZ1vRhOmq1zDRM7TjPO84P-tH5huPvI</recordid><startdate>20220207</startdate><enddate>20220207</enddate><creator>Dery, Lucio M</creator><creator>Michel, Paul</creator><creator>Talwalkar, Ameet</creator><creator>Neubig, Graham</creator><general>Cornell University Library, arXiv.org</general><scope>8FE</scope><scope>8FG</scope><scope>ABJCF</scope><scope>ABUWG</scope><scope>AFKRA</scope><scope>AZQEC</scope><scope>BENPR</scope><scope>BGLVJ</scope><scope>CCPQU</scope><scope>DWQXO</scope><scope>HCIFZ</scope><scope>L6V</scope><scope>M7S</scope><scope>PIMPY</scope><scope>PQEST</scope><scope>PQQKQ</scope><scope>PQUKI</scope><scope>PRINS</scope><scope>PTHSS</scope></search><sort><creationdate>20220207</creationdate><title>Should We Be Pre-training? An Argument for End-task Aware Training as an Alternative</title><author>Dery, Lucio M ; Michel, Paul ; Talwalkar, Ameet ; Neubig, Graham</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-proquest_journals_25731633783</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2022</creationdate><topic>Algorithms</topic><topic>Downstream effects</topic><topic>Machine learning</topic><topic>Multitasking</topic><topic>Training</topic><toplevel>online_resources</toplevel><creatorcontrib>Dery, Lucio M</creatorcontrib><creatorcontrib>Michel, Paul</creatorcontrib><creatorcontrib>Talwalkar, Ameet</creatorcontrib><creatorcontrib>Neubig, Graham</creatorcontrib><collection>ProQuest SciTech Collection</collection><collection>ProQuest Technology Collection</collection><collection>Materials Science & Engineering Collection</collection><collection>ProQuest Central (Alumni Edition)</collection><collection>ProQuest Central UK/Ireland</collection><collection>ProQuest Central Essentials</collection><collection>ProQuest Central</collection><collection>Technology Collection</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central Korea</collection><collection>SciTech Premium Collection</collection><collection>ProQuest Engineering Collection</collection><collection>Engineering Database</collection><collection>Publicly Available Content Database</collection><collection>ProQuest One Academic Eastern Edition (DO NOT USE)</collection><collection>ProQuest One Academic</collection><collection>ProQuest One Academic UKI Edition</collection><collection>ProQuest Central China</collection><collection>Engineering Collection</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Dery, Lucio M</au><au>Michel, Paul</au><au>Talwalkar, Ameet</au><au>Neubig, Graham</au><format>book</format><genre>document</genre><ristype>GEN</ristype><atitle>Should We Be Pre-training? An Argument for End-task Aware Training as an Alternative</atitle><jtitle>arXiv.org</jtitle><date>2022-02-07</date><risdate>2022</risdate><eissn>2331-8422</eissn><abstract>In most settings of practical concern, machine learning practitioners know in advance what end-task they wish to boost with auxiliary tasks. However, widely used methods for leveraging auxiliary data like pre-training and its continued-pretraining variant are end-task agnostic: they rarely, if ever, exploit knowledge of the target task. We study replacing end-task agnostic continued training of pre-trained language models with end-task aware training of said models. We argue that for sufficiently important end-tasks, the benefits of leveraging auxiliary data in a task-aware fashion can justify forgoing the traditional approach of obtaining generic, end-task agnostic representations as with (continued) pre-training. On three different low-resource NLP tasks from two domains, we demonstrate that multi-tasking the end-task and auxiliary objectives results in significantly better downstream task performance than the widely-used task-agnostic continued pre-training paradigm of Gururangan et al. (2020). We next introduce an online meta-learning algorithm that learns a set of multi-task weights to better balance among our multiple auxiliary objectives, achieving further improvements on end-task performance and data efficiency.</abstract><cop>Ithaca</cop><pub>Cornell University Library, arXiv.org</pub><oa>free_for_read</oa></addata></record> |
fulltext | fulltext |
identifier | EISSN: 2331-8422 |
ispartof | arXiv.org, 2022-02 |
issn | 2331-8422 |
language | eng |
recordid | cdi_proquest_journals_2573163378 |
source | Free E- Journals |
subjects | Algorithms Downstream effects Machine learning Multitasking Training |
title | Should We Be Pre-training? An Argument for End-task Aware Training as an Alternative |
url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-04T05%3A59%3A50IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest&rft_val_fmt=info:ofi/fmt:kev:mtx:book&rft.genre=document&rft.atitle=Should%20We%20Be%20Pre-training?%20An%20Argument%20for%20End-task%20Aware%20Training%20as%20an%20Alternative&rft.jtitle=arXiv.org&rft.au=Dery,%20Lucio%20M&rft.date=2022-02-07&rft.eissn=2331-8422&rft_id=info:doi/&rft_dat=%3Cproquest%3E2573163378%3C/proquest%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2573163378&rft_id=info:pmid/&rfr_iscdi=true |