Federated Clustering: An Overview of Algorithm Evolution and Research Prospects.

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Federated clustering (FC) emerges as a privacy-preserving paradigm for distributed data analysis.
• It integrates clustering techniques with federated learning principles.
• Addressing data privacy concerns is a key driver for FC's increasing attention.

Purpose of the Study:

• To provide a comprehensive survey of recent advancements in federated clustering.
• To systematically review classical clustering paradigms extended for FC.
• To categorize common challenges and improvement strategies in FC.

Main Methods:

• Review of classical clustering methods adapted for FC.
• Categorization of FC methods based on data partitioning and deep representation learning.
• Summary of experimental setups and evaluation protocols in FC research.
• Analysis of representative algorithms within four distinct FC categories.

Main Results:

• Identification of inherent challenges in FC and common strategies to address them.
• Classification of FC approaches into four categories based on data partitioning and deep learning integration.
• Overview of experimental methodologies and evaluation metrics used in the field.

Conclusions:

• Current FC approaches have limitations that warrant further investigation.
• Future research should explore novel FC algorithms and address identified challenges.
• The survey highlights key directions for advancing privacy-preserving distributed clustering.