Forecasting Early with Meta Learning

In the early observation period of a time series, there might be only a few historic observations available to learn a model. However, in cases where an existing prior set of datasets is available, Meta learning methods can be applicable. In this paper, we devise a Meta learning method that exploits...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Jawed, Shayan, Madhusudhanan, Kiran, Yalavarthi, Vijaya Krishna, Schmidt-Thieme, Lars
Format: Artikel
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 Jawed, Shayan
Madhusudhanan, Kiran
Yalavarthi, Vijaya Krishna
Schmidt-Thieme, Lars
description In the early observation period of a time series, there might be only a few historic observations available to learn a model. However, in cases where an existing prior set of datasets is available, Meta learning methods can be applicable. In this paper, we devise a Meta learning method that exploits samples from additional datasets and learns to augment time series through adversarial learning as an auxiliary task for the target dataset. Our model (FEML), is equipped with a shared Convolutional backbone that learns features for varying length inputs from different datasets and has dataset specific heads to forecast for different output lengths. We show that FEML can meta learn across datasets and by additionally learning on adversarial generated samples as auxiliary samples for the target dataset, it can improve the forecasting performance compared to single task learning, and various solutions adapted from Joint learning, Multi-task learning and classic forecasting baselines.
doi_str_mv 10.48550/arxiv.2307.09796
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2307_09796</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2307_09796</sourcerecordid><originalsourceid>FETCH-LOGICAL-a676-7d32c19a30c682b38d6918642e9ab21e21174a13c8eb914c482f5a6227a842923</originalsourceid><addsrcrecordid>eNotzjsPgjAYheEuDgb9AU4yuILt19LLaIy3BOPCTj5KURJEU4iXf-91OsObnDyETBiNhU4SOkf_qG8xcKpiapSRQzJbX7yz2PV1ewxX6JtneK_7U7h3PYapQ9--w4gMKmw6N_5vQLL1Kltuo_Sw2S0XaYRSyUiVHCwzyKmVGgquS2mYlgKcwQKYA8aUQMatdoVhwgoNVYISQKEWYIAHZPq7_TLzq6_P6J_5h5t_ufwFVEY4BA</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Forecasting Early with Meta Learning</title><source>arXiv.org</source><creator>Jawed, Shayan ; Madhusudhanan, Kiran ; Yalavarthi, Vijaya Krishna ; Schmidt-Thieme, Lars</creator><creatorcontrib>Jawed, Shayan ; Madhusudhanan, Kiran ; Yalavarthi, Vijaya Krishna ; Schmidt-Thieme, Lars</creatorcontrib><description>In the early observation period of a time series, there might be only a few historic observations available to learn a model. However, in cases where an existing prior set of datasets is available, Meta learning methods can be applicable. In this paper, we devise a Meta learning method that exploits samples from additional datasets and learns to augment time series through adversarial learning as an auxiliary task for the target dataset. Our model (FEML), is equipped with a shared Convolutional backbone that learns features for varying length inputs from different datasets and has dataset specific heads to forecast for different output lengths. We show that FEML can meta learn across datasets and by additionally learning on adversarial generated samples as auxiliary samples for the target dataset, it can improve the forecasting performance compared to single task learning, and various solutions adapted from Joint learning, Multi-task learning and classic forecasting baselines.</description><identifier>DOI: 10.48550/arxiv.2307.09796</identifier><language>eng</language><subject>Computer Science - Learning</subject><creationdate>2023-07</creationdate><rights>http://creativecommons.org/licenses/by/4.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,780,885</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2307.09796$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2307.09796$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Jawed, Shayan</creatorcontrib><creatorcontrib>Madhusudhanan, Kiran</creatorcontrib><creatorcontrib>Yalavarthi, Vijaya Krishna</creatorcontrib><creatorcontrib>Schmidt-Thieme, Lars</creatorcontrib><title>Forecasting Early with Meta Learning</title><description>In the early observation period of a time series, there might be only a few historic observations available to learn a model. However, in cases where an existing prior set of datasets is available, Meta learning methods can be applicable. In this paper, we devise a Meta learning method that exploits samples from additional datasets and learns to augment time series through adversarial learning as an auxiliary task for the target dataset. Our model (FEML), is equipped with a shared Convolutional backbone that learns features for varying length inputs from different datasets and has dataset specific heads to forecast for different output lengths. We show that FEML can meta learn across datasets and by additionally learning on adversarial generated samples as auxiliary samples for the target dataset, it can improve the forecasting performance compared to single task learning, and various solutions adapted from Joint learning, Multi-task learning and classic forecasting baselines.</description><subject>Computer Science - Learning</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2023</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotzjsPgjAYheEuDgb9AU4yuILt19LLaIy3BOPCTj5KURJEU4iXf-91OsObnDyETBiNhU4SOkf_qG8xcKpiapSRQzJbX7yz2PV1ewxX6JtneK_7U7h3PYapQ9--w4gMKmw6N_5vQLL1Kltuo_Sw2S0XaYRSyUiVHCwzyKmVGgquS2mYlgKcwQKYA8aUQMatdoVhwgoNVYISQKEWYIAHZPq7_TLzq6_P6J_5h5t_ufwFVEY4BA</recordid><startdate>20230719</startdate><enddate>20230719</enddate><creator>Jawed, Shayan</creator><creator>Madhusudhanan, Kiran</creator><creator>Yalavarthi, Vijaya Krishna</creator><creator>Schmidt-Thieme, Lars</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20230719</creationdate><title>Forecasting Early with Meta Learning</title><author>Jawed, Shayan ; Madhusudhanan, Kiran ; Yalavarthi, Vijaya Krishna ; Schmidt-Thieme, Lars</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a676-7d32c19a30c682b38d6918642e9ab21e21174a13c8eb914c482f5a6227a842923</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2023</creationdate><topic>Computer Science - Learning</topic><toplevel>online_resources</toplevel><creatorcontrib>Jawed, Shayan</creatorcontrib><creatorcontrib>Madhusudhanan, Kiran</creatorcontrib><creatorcontrib>Yalavarthi, Vijaya Krishna</creatorcontrib><creatorcontrib>Schmidt-Thieme, Lars</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Jawed, Shayan</au><au>Madhusudhanan, Kiran</au><au>Yalavarthi, Vijaya Krishna</au><au>Schmidt-Thieme, Lars</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Forecasting Early with Meta Learning</atitle><date>2023-07-19</date><risdate>2023</risdate><abstract>In the early observation period of a time series, there might be only a few historic observations available to learn a model. However, in cases where an existing prior set of datasets is available, Meta learning methods can be applicable. In this paper, we devise a Meta learning method that exploits samples from additional datasets and learns to augment time series through adversarial learning as an auxiliary task for the target dataset. Our model (FEML), is equipped with a shared Convolutional backbone that learns features for varying length inputs from different datasets and has dataset specific heads to forecast for different output lengths. We show that FEML can meta learn across datasets and by additionally learning on adversarial generated samples as auxiliary samples for the target dataset, it can improve the forecasting performance compared to single task learning, and various solutions adapted from Joint learning, Multi-task learning and classic forecasting baselines.</abstract><doi>10.48550/arxiv.2307.09796</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2307.09796
ispartof
issn
language eng
recordid cdi_arxiv_primary_2307_09796
source arXiv.org
subjects Computer Science - Learning
title Forecasting Early with Meta Learning
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-21T08%3A44%3A19IST&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=Forecasting%20Early%20with%20Meta%20Learning&rft.au=Jawed,%20Shayan&rft.date=2023-07-19&rft_id=info:doi/10.48550/arxiv.2307.09796&rft_dat=%3Carxiv_GOX%3E2307_09796%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