Loading…

Convolutional Neural Networks for Time-dependent Classification of Variable-length Time Series

Time series data are often obtained only within a limited time range due to interruptions during observation process. To classify such partial time series, we need to account for 1) the variable-length data drawn from 2) different timestamps. To address the first problem, existing convolutional neur...

Full description

Saved in:
Bibliographic Details
Main Authors: Sawada, Azusa, Miyagawa, Taiki, Ebihara, Akinori, Yachida, Shoji, Hosoi, Toshinori
Format: Conference Proceeding
Language:English
Subjects:
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
cited_by
cites
container_end_page 8
container_issue
container_start_page 1
container_title
container_volume
creator Sawada, Azusa
Miyagawa, Taiki
Ebihara, Akinori
Yachida, Shoji
Hosoi, Toshinori
description Time series data are often obtained only within a limited time range due to interruptions during observation process. To classify such partial time series, we need to account for 1) the variable-length data drawn from 2) different timestamps. To address the first problem, existing convolutional neural networks use global pooling after convolutional layers to cancel the length differences. This architecture suffers from the trade-off between incorporating entire temporal correlations in long data and avoiding feature collapse for short data. To resolve this trade-off, we propose Adaptive Multi-scale Pooling, which aggregates features from an adaptive number of layers, i.e., only the first few layers for short data and more layers for long data. Furthermore, to address the second problem, we introduce Temporal Encoding, which embeds the observation timestamps into the intermediate features. Experiments on our private dataset and the UCR/UEA time series archive show that our modules improve classification accuracy especially on short data obtained as partial time series.
doi_str_mv 10.1109/IJCNN55064.2022.9892605
format conference_proceeding
fullrecord <record><control><sourceid>ieee_CHZPO</sourceid><recordid>TN_cdi_ieee_primary_9892605</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><ieee_id>9892605</ieee_id><sourcerecordid>9892605</sourcerecordid><originalsourceid>FETCH-LOGICAL-i203t-5ba8461a8b710df98785eb78d96744a15c36bab648bdd149e69f9b47281145f43</originalsourceid><addsrcrecordid>eNotkM1OAjEURquJiYA-gQv7AoO9nf4uzUQFQ8aF6FLSMrdaHWZIO2h8ewOyOqvzJecj5BrYFIDZm_ljVddSMiWmnHE-tcZyxeQJGYPmBozSIE7JiIOCQgimz8k450_GeGltOSJvVd999-1uiH3nWlrjLh0w_PTpK9PQJ7qMGywa3GLXYDfQqnU5xxDXbu_QPtBXl6LzLRYtdu_Dx0Ggz5gi5gtyFlyb8fLICXm5v1tWs2Lx9DCvbhdF5KwcCumdEQqc8RpYE6zRRqLXprFKC-FArkvlnVfC-KYBYVHZYL3YB4KQQZQTcvW_GxFxtU1x49Lv6vhF-Qe_3lUw</addsrcrecordid><sourcetype>Publisher</sourcetype><iscdi>true</iscdi><recordtype>conference_proceeding</recordtype></control><display><type>conference_proceeding</type><title>Convolutional Neural Networks for Time-dependent Classification of Variable-length Time Series</title><source>IEEE Xplore All Conference Series</source><creator>Sawada, Azusa ; Miyagawa, Taiki ; Ebihara, Akinori ; Yachida, Shoji ; Hosoi, Toshinori</creator><creatorcontrib>Sawada, Azusa ; Miyagawa, Taiki ; Ebihara, Akinori ; Yachida, Shoji ; Hosoi, Toshinori</creatorcontrib><description>Time series data are often obtained only within a limited time range due to interruptions during observation process. To classify such partial time series, we need to account for 1) the variable-length data drawn from 2) different timestamps. To address the first problem, existing convolutional neural networks use global pooling after convolutional layers to cancel the length differences. This architecture suffers from the trade-off between incorporating entire temporal correlations in long data and avoiding feature collapse for short data. To resolve this trade-off, we propose Adaptive Multi-scale Pooling, which aggregates features from an adaptive number of layers, i.e., only the first few layers for short data and more layers for long data. Furthermore, to address the second problem, we introduce Temporal Encoding, which embeds the observation timestamps into the intermediate features. Experiments on our private dataset and the UCR/UEA time series archive show that our modules improve classification accuracy especially on short data obtained as partial time series.</description><identifier>EISSN: 2161-4407</identifier><identifier>EISBN: 1728186714</identifier><identifier>EISBN: 9781728186719</identifier><identifier>DOI: 10.1109/IJCNN55064.2022.9892605</identifier><language>eng</language><publisher>IEEE</publisher><subject>Aggregates ; convolutional neural networks ; Correlation ; Encoding ; Neural networks ; partial data ; Radio frequency ; Time series analysis ; time series classification ; Trajectory ; variable length</subject><ispartof>2022 International Joint Conference on Neural Networks (IJCNN), 2022, p.1-8</ispartof><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://ieeexplore.ieee.org/document/9892605$$EHTML$$P50$$Gieee$$H</linktohtml><link.rule.ids>309,310,780,784,789,790,23929,23930,25139,27924,54554,54931</link.rule.ids><linktorsrc>$$Uhttps://ieeexplore.ieee.org/document/9892605$$EView_record_in_IEEE$$FView_record_in_$$GIEEE</linktorsrc></links><search><creatorcontrib>Sawada, Azusa</creatorcontrib><creatorcontrib>Miyagawa, Taiki</creatorcontrib><creatorcontrib>Ebihara, Akinori</creatorcontrib><creatorcontrib>Yachida, Shoji</creatorcontrib><creatorcontrib>Hosoi, Toshinori</creatorcontrib><title>Convolutional Neural Networks for Time-dependent Classification of Variable-length Time Series</title><title>2022 International Joint Conference on Neural Networks (IJCNN)</title><addtitle>IJCNN</addtitle><description>Time series data are often obtained only within a limited time range due to interruptions during observation process. To classify such partial time series, we need to account for 1) the variable-length data drawn from 2) different timestamps. To address the first problem, existing convolutional neural networks use global pooling after convolutional layers to cancel the length differences. This architecture suffers from the trade-off between incorporating entire temporal correlations in long data and avoiding feature collapse for short data. To resolve this trade-off, we propose Adaptive Multi-scale Pooling, which aggregates features from an adaptive number of layers, i.e., only the first few layers for short data and more layers for long data. Furthermore, to address the second problem, we introduce Temporal Encoding, which embeds the observation timestamps into the intermediate features. Experiments on our private dataset and the UCR/UEA time series archive show that our modules improve classification accuracy especially on short data obtained as partial time series.</description><subject>Aggregates</subject><subject>convolutional neural networks</subject><subject>Correlation</subject><subject>Encoding</subject><subject>Neural networks</subject><subject>partial data</subject><subject>Radio frequency</subject><subject>Time series analysis</subject><subject>time series classification</subject><subject>Trajectory</subject><subject>variable length</subject><issn>2161-4407</issn><isbn>1728186714</isbn><isbn>9781728186719</isbn><fulltext>true</fulltext><rsrctype>conference_proceeding</rsrctype><creationdate>2022</creationdate><recordtype>conference_proceeding</recordtype><sourceid>6IE</sourceid><recordid>eNotkM1OAjEURquJiYA-gQv7AoO9nf4uzUQFQ8aF6FLSMrdaHWZIO2h8ewOyOqvzJecj5BrYFIDZm_ljVddSMiWmnHE-tcZyxeQJGYPmBozSIE7JiIOCQgimz8k450_GeGltOSJvVd999-1uiH3nWlrjLh0w_PTpK9PQJ7qMGywa3GLXYDfQqnU5xxDXbu_QPtBXl6LzLRYtdu_Dx0Ggz5gi5gtyFlyb8fLICXm5v1tWs2Lx9DCvbhdF5KwcCumdEQqc8RpYE6zRRqLXprFKC-FArkvlnVfC-KYBYVHZYL3YB4KQQZQTcvW_GxFxtU1x49Lv6vhF-Qe_3lUw</recordid><startdate>20220718</startdate><enddate>20220718</enddate><creator>Sawada, Azusa</creator><creator>Miyagawa, Taiki</creator><creator>Ebihara, Akinori</creator><creator>Yachida, Shoji</creator><creator>Hosoi, Toshinori</creator><general>IEEE</general><scope>6IE</scope><scope>6IH</scope><scope>CBEJK</scope><scope>RIE</scope><scope>RIO</scope></search><sort><creationdate>20220718</creationdate><title>Convolutional Neural Networks for Time-dependent Classification of Variable-length Time Series</title><author>Sawada, Azusa ; Miyagawa, Taiki ; Ebihara, Akinori ; Yachida, Shoji ; Hosoi, Toshinori</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-i203t-5ba8461a8b710df98785eb78d96744a15c36bab648bdd149e69f9b47281145f43</frbrgroupid><rsrctype>conference_proceedings</rsrctype><prefilter>conference_proceedings</prefilter><language>eng</language><creationdate>2022</creationdate><topic>Aggregates</topic><topic>convolutional neural networks</topic><topic>Correlation</topic><topic>Encoding</topic><topic>Neural networks</topic><topic>partial data</topic><topic>Radio frequency</topic><topic>Time series analysis</topic><topic>time series classification</topic><topic>Trajectory</topic><topic>variable length</topic><toplevel>online_resources</toplevel><creatorcontrib>Sawada, Azusa</creatorcontrib><creatorcontrib>Miyagawa, Taiki</creatorcontrib><creatorcontrib>Ebihara, Akinori</creatorcontrib><creatorcontrib>Yachida, Shoji</creatorcontrib><creatorcontrib>Hosoi, Toshinori</creatorcontrib><collection>IEEE Electronic Library (IEL) Conference Proceedings</collection><collection>IEEE Proceedings Order Plan (POP) 1998-present by volume</collection><collection>IEEE Xplore All Conference Proceedings</collection><collection>IEEE</collection><collection>IEEE Proceedings Order Plans (POP) 1998-present</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Sawada, Azusa</au><au>Miyagawa, Taiki</au><au>Ebihara, Akinori</au><au>Yachida, Shoji</au><au>Hosoi, Toshinori</au><format>book</format><genre>proceeding</genre><ristype>CONF</ristype><atitle>Convolutional Neural Networks for Time-dependent Classification of Variable-length Time Series</atitle><btitle>2022 International Joint Conference on Neural Networks (IJCNN)</btitle><stitle>IJCNN</stitle><date>2022-07-18</date><risdate>2022</risdate><spage>1</spage><epage>8</epage><pages>1-8</pages><eissn>2161-4407</eissn><eisbn>1728186714</eisbn><eisbn>9781728186719</eisbn><abstract>Time series data are often obtained only within a limited time range due to interruptions during observation process. To classify such partial time series, we need to account for 1) the variable-length data drawn from 2) different timestamps. To address the first problem, existing convolutional neural networks use global pooling after convolutional layers to cancel the length differences. This architecture suffers from the trade-off between incorporating entire temporal correlations in long data and avoiding feature collapse for short data. To resolve this trade-off, we propose Adaptive Multi-scale Pooling, which aggregates features from an adaptive number of layers, i.e., only the first few layers for short data and more layers for long data. Furthermore, to address the second problem, we introduce Temporal Encoding, which embeds the observation timestamps into the intermediate features. Experiments on our private dataset and the UCR/UEA time series archive show that our modules improve classification accuracy especially on short data obtained as partial time series.</abstract><pub>IEEE</pub><doi>10.1109/IJCNN55064.2022.9892605</doi><tpages>8</tpages></addata></record>
fulltext fulltext_linktorsrc
identifier EISSN: 2161-4407
ispartof 2022 International Joint Conference on Neural Networks (IJCNN), 2022, p.1-8
issn 2161-4407
language eng
recordid cdi_ieee_primary_9892605
source IEEE Xplore All Conference Series
subjects Aggregates
convolutional neural networks
Correlation
Encoding
Neural networks
partial data
Radio frequency
Time series analysis
time series classification
Trajectory
variable length
title Convolutional Neural Networks for Time-dependent Classification of Variable-length Time Series
url http://sfxeu10.hosted.exlibrisgroup.com/loughborough?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-12T23%3A13%3A36IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-ieee_CHZPO&rft_val_fmt=info:ofi/fmt:kev:mtx:book&rft.genre=proceeding&rft.atitle=Convolutional%20Neural%20Networks%20for%20Time-dependent%20Classification%20of%20Variable-length%20Time%20Series&rft.btitle=2022%20International%20Joint%20Conference%20on%20Neural%20Networks%20(IJCNN)&rft.au=Sawada,%20Azusa&rft.date=2022-07-18&rft.spage=1&rft.epage=8&rft.pages=1-8&rft.eissn=2161-4407&rft_id=info:doi/10.1109/IJCNN55064.2022.9892605&rft.eisbn=1728186714&rft.eisbn_list=9781728186719&rft_dat=%3Cieee_CHZPO%3E9892605%3C/ieee_CHZPO%3E%3Cgrp_id%3Ecdi_FETCH-LOGICAL-i203t-5ba8461a8b710df98785eb78d96744a15c36bab648bdd149e69f9b47281145f43%3C/grp_id%3E%3Coa%3E%3C/oa%3E%3Curl%3E%3C/url%3E&rft_id=info:oai/&rft_id=info:pmid/&rft_ieee_id=9892605&rfr_iscdi=true