Scholarlink

The growing interest in federated learning (FL) has led to increased attention on hyperparameter optimization (HPO) within distributed settings. However, effectively applying HPO in practical FL scenarios presents unique challenges owing to the decentralized characteristics of data and computation. This study provides a comprehensive examination of various HPO strategies tailored for FL, addressing the limitations of earlier analyses that primarily focused on centralized environments with unrestricted data access and consolidated processing resources. In contrast, FL imposes constraints where individual participants operate on isolated local datasets without sharing raw data. As a result, conventional HPO techniques must be adapted to accommodate these decentralized conditions. We explore these adaptations and highlight promising directions for future work. Furthermore, we evaluate the performance of Random Search, a robust yet straightforward HPO approach, through empirical testing in a realistic federated setup using the MNIST dataset in a non-IID setting.

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

Analyzing Hyperparameter Optimization Methods for Federated Learning Systems

  • 会议论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年

查看全文
  • Sa Jim Soe Moe,
  • Qazi Waqas Khan,
  • Nguyen Anh Tuan,
  • Misbah Bibi,
  • Dohyeun Kim

摘要

The growing interest in federated learning (FL) has led to increased attention on hyperparameter optimization (HPO) within distributed settings. However, effectively applying HPO in practical FL scenarios presents unique challenges owing to the decentralized characteristics of data and computation. This study provides a comprehensive examination of various HPO strategies tailored for FL, addressing the limitations of earlier analyses that primarily focused on centralized environments with unrestricted data access and consolidated processing resources. In contrast, FL imposes constraints where individual participants operate on isolated local datasets without sharing raw data. As a result, conventional HPO techniques must be adapted to accommodate these decentralized conditions. We explore these adaptations and highlight promising directions for future work. Furthermore, we evaluate the performance of Random Search, a robust yet straightforward HPO approach, through empirical testing in a realistic federated setup using the MNIST dataset in a non-IID setting.