Scholarlink

<p>The rapid increase in data produced in smart environments offers substantial potential for applying machine learning to enhance decision-making and boost efficiency. Traditional machine learning methods struggle with distributed datasets and privacy issues. Federated learning presents a novel solution that allows collaborative learning while protecting privacy. This paper presents Smart Federated Middleware for Educational Institutions (SFMEI), a middleware aimed at supporting application development that uses federated learning within smart campuses. It also facilitates inter-campus collaboration for building models with privately distributed data. Experiments show the effectiveness of SFMEI in maintaining predictive performance while protecting data privacy. We used an Long Short-Term Memory (LSTM) network to forecast time series data on CO2 levels, temperature, energy use, and water consumption. SFMEI demonstrated efficacy, attaining an <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="12652_2025_4978_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="21" /> </InlineMediaObject> <EquationSource Format="TEX">\(R^2\)</EquationSource> </InlineEquation> score comparable to that in a centralized environment. Furthermore, the study compared different federated learning aggregation algorithms, specifically FedAVG and FedSGD, with FedSGD outperforming FedAVG in most cases.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Federated learning middleware for smart campus solutions

  • 期刊论文

  • 发表日期：

    2025年5月24日

查看全文
  • Lucas E. B. dos Santos,
  • Paulo R. Lins Júnior,
  • Ruan D. Gomes

摘要

The rapid increase in data produced in smart environments offers substantial potential for applying machine learning to enhance decision-making and boost efficiency. Traditional machine learning methods struggle with distributed datasets and privacy issues. Federated learning presents a novel solution that allows collaborative learning while protecting privacy. This paper presents Smart Federated Middleware for Educational Institutions (SFMEI), a middleware aimed at supporting application development that uses federated learning within smart campuses. It also facilitates inter-campus collaboration for building models with privately distributed data. Experiments show the effectiveness of SFMEI in maintaining predictive performance while protecting data privacy. We used an Long Short-Term Memory (LSTM) network to forecast time series data on CO2 levels, temperature, energy use, and water consumption. SFMEI demonstrated efficacy, attaining an \(R^2\) score comparable to that in a centralized environment. Furthermore, the study compared different federated learning aggregation algorithms, specifically FedAVG and FedSGD, with FedSGD outperforming FedAVG in most cases.