Scholarlink

<p>This study presents a bibliometric and thematic analysis of Personalized Federated Learning (PFL) research, synthesizing a curated corpus of 1,204 unique publications from Scopus, Web of Science, and IEEE Xplore (2020–2025). Employing logistic growth modeling, network centrality analysis, and interpretable topic extraction, we provide a data-driven mapping of the field’s evolution and intellectual structure. Quantitative analysis reveals exponential growth (68% CAGR), with logistic modeling projecting peak publication activity around 2029 (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(t_0 = 2029.3\)</EquationSource> </InlineEquation>, <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(K = 11,120\)</EquationSource> </InlineEquation>, <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(R^2 = 0.971\)</EquationSource> </InlineEquation>). Network analysis identifies six thematic clusters, with “federated learning”, “personalization”, and “data heterogeneity” as primary conceptual hubs. Country-level collaboration analysis shows China (274 documents), the United States (113), and the United Kingdom (78) as primary hubs, while Switzerland, Singapore, and the UAE occupy critical brokerage positions. Topic modeling reveals four dominant thematic pillars; fairness (107 documents), trust frameworks (190), blockchain transparency (178), and robustness (302). The analysis identifies critical gaps including holistic trust mechanisms, adversarial resilience, and equitable client incentivization. Based on thematic evolution and gap analysis, we outline literature-grounded future directions including bias-aware personalization, scalable accountability architectures, and context-aware aggregation for high-stakes applications in healthcare, finance, and edge intelligence.</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

Mapping the Intellectual Landscape of Personalized Federated Learning: A Bibliometric Analysis

  • 期刊论文

  • 发表日期：

    2026年7月24日

查看全文
  • Fahad Sabah,
  • Yuwen Chen,
  • Yang Zhen,
  • Muhammad Azam,
  • Nadeem Ahmad,
  • Raheem Sarwar

摘要

This study presents a bibliometric and thematic analysis of Personalized Federated Learning (PFL) research, synthesizing a curated corpus of 1,204 unique publications from Scopus, Web of Science, and IEEE Xplore (2020–2025). Employing logistic growth modeling, network centrality analysis, and interpretable topic extraction, we provide a data-driven mapping of the field’s evolution and intellectual structure. Quantitative analysis reveals exponential growth (68% CAGR), with logistic modeling projecting peak publication activity around 2029 ( \(t_0 = 2029.3\) , \(K = 11,120\) , \(R^2 = 0.971\) ). Network analysis identifies six thematic clusters, with “federated learning”, “personalization”, and “data heterogeneity” as primary conceptual hubs. Country-level collaboration analysis shows China (274 documents), the United States (113), and the United Kingdom (78) as primary hubs, while Switzerland, Singapore, and the UAE occupy critical brokerage positions. Topic modeling reveals four dominant thematic pillars; fairness (107 documents), trust frameworks (190), blockchain transparency (178), and robustness (302). The analysis identifies critical gaps including holistic trust mechanisms, adversarial resilience, and equitable client incentivization. Based on thematic evolution and gap analysis, we outline literature-grounded future directions including bias-aware personalization, scalable accountability architectures, and context-aware aggregation for high-stakes applications in healthcare, finance, and edge intelligence.