Federated Learning Study Reveals Top Developer Pain Points Across 92 Projects

A new empirical study titled "Understanding Developer Pain Points in Federated Learning: Insights from Stack Overflow and GitHub" systematically maps the challenges developers face when building and deploying federated learning (FL) systems. The authors, Sahand Saed, Khairul Alam, and Banani Roy, analyzed 495 Stack Overflow posts and 9,116 GitHub issues and pull requests from 92 FL-related open-source projects. Using BERTopic-based topic modeling and difficulty indicators such as unresolved rates and median resolution time, they surfaced nine dominant topics on Stack Overflow and thirteen on GitHub.

The recurring pain points include environment setup and dependency compatibility, API breakages and migration issues, training instability under non-IID data distributions, evaluation and metric correctness, and the integration of privacy-preserving mechanisms. A key finding is that "How"-type questions dominate, indicating a strong need for procedural guidance. Specific topics like "TFF Installation and Environment Compatibility" and "Federated Feature Engineering and SecureBoost Issues" showed high unresolved rates and long resolution times, pointing to gaps in tooling, documentation, and debugging support. The researchers conclude with actionable implications for framework designers, documentation authors, and educators to improve FL usability and deployability.