Personalized Progressive Federated Learning with Leveraging Client-Specific Vertical Features

<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"> <html xmlns="http://www.w3.org/1999/xhtml"> <head> <title>Personalized Progressive Federated Learning with Leveraging Client-Specific Vertical Features</title>  <meta charset="UTF-8"> <meta name="viewport" content="width=device-width, initial-scale=1.0"> <meta http-equiv="X-UA-Compatible" content="ie=edge"> <meta name="title" content="Personalized Progressive Federated Learning with Leveraging Client-Specific Vertical Features"> <meta name="description" content="Federated learning (FL) has been used for model building across distributed clients. However, FL cannot leverage vertically partitioned features to increase the model complexity. In this study, we proposed a personalized progressive federated learning (PPFL) model, which is a multimodel PFL approach that allows the leveraging of vertically partitioned client-specific features. The performance of PPFL was evaluated using the Physionet Challenges 2012 dataset. We compared the performance of in-hospital mortality and length of stay prediction between our model and the FedAvg, FedProx, and local models. The PPFL showed an accuracy of 0.849 and AUROC of 0.790 in average in hospital mor-tality prediction, which are the highest scores compared to client-specific algorithm. For length-of-stay prediction, PPFL also showed an AUROC of 0.808 in average which was the highest among all comparators"/> <meta name="keywords" content="Personalized Federated Learning, Vertical Federated Learning, Non-IID data"/>   <meta name="dc.title" content="Personalized Progressive Federated Learning with Leveraging Client-Specific Vertical Features"> <meta name="citation_authors" content="Tae Hyun Kim "> <meta name="citation_authors" content="Won Seok Jang"> <meta name="citation_authors" content="Sun Cheol Heo "> <meta name="citation_authors" content="MinDong Sung"> <meta name="citation_authors" content="Yu Rang Park"> <meta name="dc.type" content="Article"> <meta name="dc.source" content="Computer Science & Information Technology (CS & IT) Vol. 13, No.1"> <meta name="dc.date" content="2023/01/02"> <meta name="dc.identifier" content="10.5121/csit.2023.130106"> <meta name="dc.publisher" content="AIRCC Publishing Corporation"> <meta name="dc.rights" content="http://creativecommons.org/licenses/by/3.0/"> <meta name="dc.format" content="application/pdf"> <meta name="dc.language" content="en"> <meta name="dc.description" content="Federated learning (FL) has been used for model building across distributed clients. However, FL cannot leverage vertically partitioned features to increase the model complexity. In this study, we proposed a personalized progressive federated learning (PPFL) model, which is a multimodel PFL approach that allows the leveraging of vertically partitioned client-specific features. The performance of PPFL was evaluated using the Physionet Challenges 2012 dataset. We compared the performance of in-hospital mortality and length of stay prediction between our model and the FedAvg, FedProx, and local models. The PPFL showed an accuracy of 0.849 and AUROC of 0.790 in average in hospital mor-tality prediction, which are the highest scores compared to client-specific algorithm. For length-of-stay prediction, PPFL also showed an AUROC of 0.808 in average which was the highest among all comparators."/> <meta name="dc.subject" content="Personalized Federated Learning"> <meta name="dc.subject" content="Vertical Federated Learning"> <meta name="dc.subject" content="Non-IID data">   <meta name="prism.publicationName" content="Computer Science & Information Technology (CS & IT) "> <meta name="prism.publicationDate" content="2023/01/02"> <meta name="prism.volume" content="13"> <meta name="prism.number" content="1"> <meta name="prism.section" content="Article"> <meta name="prism.startingPage" content="65">   <meta name="citation_journal_title" content="Computer Science & Information Technology (CS & IT)"> <meta name="citation_publisher" content="AIRCC Publishing Corporation"> <meta name="citation_authors" content="Tae Hyun Kim, Won Seok Jang, Sun Cheol Heo, MinDong Sung, and Yu Rang Park"> <meta name="citation_title" content="Personalized Progressive Federated Learning with Leveraging Client-Specific Vertical Features "> <meta name="citation_online_date" content="2023/01/02"> <meta name="citation_issue" content="13"> <meta name="citation_firstpage" content="65"> <meta name="citation_authors" content="Tae Hyun Kim "> <meta name="citation_authors" content="Won Seok Jang"> <meta name="citation_authors" content="Sun Cheol Heo "> <meta name="citation_authors" content="MinDong Sung"> <meta name="citation_authors" content="Yu Rang Park"> <meta name="citation_doi" content="10.5121/csit.2023.130106"> <meta name="citation_abstract_html_url" content="https://aircconline.com/csit/abstract/v13n1/csit130106.html"> <meta name="citation_pdf_url" content="https://aircconline.com/csit/papers/vol13/csit130106.pdf">   <meta property="og:site_name" content="AIRCC" /> <meta property="og:type" content="article" /> <meta property="og:url" content="https://aircconline.com/csit/abstract/v13n1/csit130106.html"> <meta property="og:title" content="Personalized Progressive Federated Learning with Leveraging Client-Specific Vertical Features "> <meta property="og:description" content="Federated learning (FL) has been used for model building across distributed clients. However, FL cannot leverage vertically partitioned features to increase the model complexity. In this study, we proposed a personalized progressive federated learning (PPFL) model, which is a multimodel PFL approach that allows the leveraging of vertically partitioned client-specific features. The performance of PPFL was evaluated using the Physionet Challenges 2012 dataset. We compared the performance of in-hospital mortality and length of stay prediction between our model and the FedAvg, FedProx, and local models. The PPFL showed an accuracy of 0.849 and AUROC of 0.790 in average in hospital mor-tality prediction, which are the highest scores compared to client-specific algorithm. For length-of-stay prediction, PPFL also showed an AUROC of 0.808 in average which was the highest among all comparators."/>   <meta name="google-site-verification" content="t8rHIcM8EfjIqfQzQ0IdYIiA9JxDD0uUZAitBCzsOIw" /> <meta name="yandex-verification" content="e3d2d5a32c7241f4" />  <link rel="icon" type="image/ico" href="../img/ico.ico"/> <link rel="stylesheet" type="text/css" href="../main1.css" media="screen" /> <style type="text/css"> a{ color:white; text-decoration:none; line-height:20px; } ul li a{ font-weight:bold; color:#000; list-style:none; text-decoration:none; size:10px;} .imagess { height:90px; text-align:left; margin:0px 5px 2px 8px; float:right; border:none; } #left p { font-family: CALIBRI; font-size: 16px; margin-left: 20px; font-weight: 500; } .right { margin-right: 20px; } #button{ float: left; font-size: 14px; margin-left: 10px; height: 28px; width: auto; background-color: #1e86c6; } </style> </head> <body> <div class="font"> <div id="wap"> <div id="page"> <div id="top"> <form action="https://airccj.org/csecfp/library/Search.php" method="get" target="_blank" > <table width="100%" cellspacing="0" cellpadding="0" > <tr class="search_input"> <td width="665" align="right"> </td> <td width="236" > <input name="title" type="text" value="Enter the paper title" class="search_textbox" onclick="if(this.value=='Enter the paper title'){this.value=''}" onblur="if(this.value==''){this.value='Enter the paper title'}" /> </td> <td width="59"> <input type="image" src="../img/go.gif" /> </td> </tr> <tr> <td colspan="3" valign="top"><img src="../img/top1.gif" alt="Academy & Industry Research Collaboration Center (AIRCC)" /></td> </tr> </table> </form> </div> <div id="font-face"> <div id="menu"> <a href="http://airccse.org">Home</a> <a href="http://airccse.org/journal.html">Journals</a> <a href="http://airccse.org/ethics.html">Ethics</a> <a href="http://airccse.org/conference.html">Conferences</a> <a href="http://airccse.org/past.html">Past Events</a> <a href="http://airccse.org/b.html">Submission</a> </div> <div id="content"> <div id="left"> <h2 class="lighter"><font size="2">Volume 13, Number 01, January 2023</font></h2> <h4 style="text-align:center;height:auto;"><a>Personalized Progressive Federated Learning with Leveraging Client-Specific Vertical Features </a></h4> <h3>  Authors</h3> <p class="#left right" style="text-align:">Tae Hyun Kim, Won Seok Jang, Sun Cheol Heo, MinDong Sung, and Yu Rang Park, Yonsei University College of Medicine, South Korea </p> <h3>  Abstract</h3> <p class="#left right" style="text-align:justify">Federated learning (FL) has been used for model building across distributed clients. However, FL cannot leverage vertically partitioned features to increase the model complexity. In this study, we proposed a personalized progressive federated learning (PPFL) model, which is a multimodel PFL approach that allows the leveraging of vertically partitioned client-specific features. The performance of PPFL was evaluated using the Physionet Challenges 2012 dataset. We compared the performance of in-hospital mortality and length of stay prediction between our model and the FedAvg, FedProx, and local models. The PPFL showed an accuracy of 0.849 and AUROC of 0.790 in average in hospital mor-tality prediction, which are the highest scores compared to client-specific algorithm. For length-of-stay prediction, PPFL also showed an AUROC of 0.808 in average which was the highest among all comparators. </p> <h3>  Keywords</h3> <p class="#left right" style="text-align:justify">Personalized Federated Learning, Vertical Federated Learning, Non-IID data.</p><br> <button type="button" id="button"><a target="_blank" href="/csit/papers/vol13/csit130106.pdf">Full Text</a></button>   <button type="button" id="button"><a href="http://airccse.org/csit/V13N01.html">Volume 13, Number 01</a></button> <br><br><br><br><br> </div> <div id="right"> <div class="menu_right"> <ul> <li id="id"><a href="http://airccse.org/editorial.html">Editorial Board</a></li> <li><a href="http://airccse.org/arch.html">Archives</a></li> <li><a href="http://airccse.org/indexing.html">Indexing</a></li> <li><a href="http://airccse.org/faq.html" target="_blank">FAQ</a></li> </ul> </div> <div class="clear_left"></div> <br> </div> <div class="clear"></div> <div id="footer"> <table width="100%" > <tr> <td width="46%" class="F_menu"><a href="http://airccse.org/subscription.html">Subscription</a> <a href="http://airccse.org/membership.html">Membership</a> <a href="http://airccse.org/cscp.html">AIRCC CSCP</a> <a href="http://airccse.org/acontact.html">Contact Us</a> </td> <td width="54%" align="right"><a href="http://airccse.org/index.php"><img src="/csit/abstract/img/logo.gif" alt="" width="21" height="24" /></a><a href="http://www.facebook.com/AIRCCSE"><img src="/csit/abstract/img/facebook.jpeg" alt="" width="21" height="24" /></a><a href="https://twitter.com/AIRCCFP"><img src="/csit/abstract/img/twitter.jpeg" alt="" width="21" height="24" /></a><a href="http://cfptech.wordpress.com/"><img src="/csit/abstract/img/index1.jpeg" alt="" width="21" height="24" /></a></td> </tr> <tr><td height="25" colspan="2"> <p align="center">All Rights Reserved ® AIRCC</p> </td></tr> </table> </div> </div> </div> </div> </div> </div> </body> </html>

CINXE.COM

Personalized Progressive Federated Learning with Leveraging Client-Specific Vertical Features