Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DAMe: Personalized Federated Social Event Detection with Dual Aggregation Mechanism

Xiaoyan Yu, Yifan Wei, Pu Li, Shuaishuai Zhou, Hao Peng, Li Sun, Liehuang Zhu, Philip S. Yu

TL;DR

This paper addresses the challenge of federated social event detection under heterogeneous, multilingual data by introducing DAMe, a dual aggregation framework that personalizes local models while leveraging global knowledge. DAMe combines Bayesian optimization for local aggregation with a 2D structural entropy-based global aggregation and a global-local event-centric constraint to align representations, reducing client-drift and overfitting. The approach is validated on six multilingual datasets across two platforms, showing superior local performance relative to strong FL baselines and robustness to injection attacks. The results suggest that DAMe provides a practical and scalable solution for FedSED with strong performance, robustness, and interpretable aggregation mechanisms that can extend to other privacy-preserving, heterogeneous learning tasks.

Abstract

Training social event detection models through federated learning (FedSED) aims to improve participants' performance on the task. However, existing federated learning paradigms are inadequate for achieving FedSED's objective and exhibit limitations in handling the inherent heterogeneity in social data. This paper proposes a personalized federated learning framework with a dual aggregation mechanism for social event detection, namely DAMe. We present a novel local aggregation strategy utilizing Bayesian optimization to incorporate global knowledge while retaining local characteristics. Moreover, we introduce a global aggregation strategy to provide clients with maximum external knowledge of their preferences. In addition, we incorporate a global-local event-centric constraint to prevent local overfitting and ``client-drift''. Experiments within a realistic simulation of a natural federated setting, utilizing six social event datasets spanning six languages and two social media platforms, along with an ablation study, have demonstrated the effectiveness of the proposed framework. Further robustness analyses have shown that DAMe is resistant to injection attacks.

DAMe: Personalized Federated Social Event Detection with Dual Aggregation Mechanism

TL;DR

This paper addresses the challenge of federated social event detection under heterogeneous, multilingual data by introducing DAMe, a dual aggregation framework that personalizes local models while leveraging global knowledge. DAMe combines Bayesian optimization for local aggregation with a 2D structural entropy-based global aggregation and a global-local event-centric constraint to align representations, reducing client-drift and overfitting. The approach is validated on six multilingual datasets across two platforms, showing superior local performance relative to strong FL baselines and robustness to injection attacks. The results suggest that DAMe provides a practical and scalable solution for FedSED with strong performance, robustness, and interpretable aggregation mechanisms that can extend to other privacy-preserving, heterogeneous learning tasks.

Abstract

Training social event detection models through federated learning (FedSED) aims to improve participants' performance on the task. However, existing federated learning paradigms are inadequate for achieving FedSED's objective and exhibit limitations in handling the inherent heterogeneity in social data. This paper proposes a personalized federated learning framework with a dual aggregation mechanism for social event detection, namely DAMe. We present a novel local aggregation strategy utilizing Bayesian optimization to incorporate global knowledge while retaining local characteristics. Moreover, we introduce a global aggregation strategy to provide clients with maximum external knowledge of their preferences. In addition, we incorporate a global-local event-centric constraint to prevent local overfitting and ``client-drift''. Experiments within a realistic simulation of a natural federated setting, utilizing six social event datasets spanning six languages and two social media platforms, along with an ablation study, have demonstrated the effectiveness of the proposed framework. Further robustness analyses have shown that DAMe is resistant to injection attacks.
Paper Structure (35 sections, 18 equations, 4 figures, 4 tables)

This paper contains 35 sections, 18 equations, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Social Event Detection
  4. Federated Learning
  5. Preliminaries
  6. Federated Learning
  7. Social Event Detection
  8. Threat Model
  9. Methodology
  10. Backbone Social Event Detection Model
  11. Social Message Graph Construction
  12. Social Message Representation
  13. Local Aggregation via Bayesian Optimization
  14. Objective Function
  15. Bayesian statistical model
  16. ...and 20 more sections

Figures (4)

  • Figure 1: The overall framework of DAMe.
  • Figure 2: The results of ablation study on all datasets.
  • Figure 3: The NMI score corresponds to the aggregation weight within the BOLA search space. The overall search space is delineated by the two black lines, while the blue box represents 50% of the NMI scores associated with the search space, and the yellow line denotes the midpoint. The red dotted horizontal line illustrates the performance of DAMe without BOLA.
  • Figure 4: The convergence plots of all methods.