Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A new type of federated clustering: A non-model-sharing approach

Yuji Kawamata, Kaoru Kamijo, Masateru Kihira, Akihiro Toyoda, Tomoru Nakayama, Akira Imakura, Tetsuya Sakurai, Yukihiko Okada

TL;DR

DC-Clustering presents a privacy-preserving, non-model-sharing federated clustering framework capable of handling complex horizontal-vertical data partitions. It uses an anchor-guided construction of intermediate representations and a collaborative representation to perform either k-means or spectral clustering in a single round of communication with a central analyst. Empirical results on synthetic and open datasets show clustering performance close to centralized pooling, addressing a key gap in federated clustering for mixed data distributions. The approach offers practical value for privacy-sensitive domains by enabling knowledge discovery from distributed heterogeneous data while reducing communication overhead.

Abstract

In recent years, the growing need to leverage sensitive data across institutions has led to increased attention on federated learning (FL), a decentralized machine learning paradigm that enables model training without sharing raw data. However, existing FL-based clustering methods, known as federated clustering, typically assume simple data partitioning scenarios such as horizontal or vertical splits, and cannot handle more complex distributed structures. This study proposes data collaboration clustering (DC-Clustering), a novel federated clustering method that supports clustering over complex data partitioning scenarios where horizontal and vertical splits coexist. In DC-Clustering, each institution shares only intermediate representations instead of raw data, ensuring privacy preservation while enabling collaborative clustering. The method allows flexible selection between k-means and spectral clustering, and achieves final results with a single round of communication with the central server. We conducted extensive experiments using synthetic and open benchmark datasets. The results show that our method achieves clustering performance comparable to centralized clustering where all data are pooled. DC-Clustering addresses an important gap in current FL research by enabling effective knowledge discovery from distributed heterogeneous data. Its practical properties -- privacy preservation, communication efficiency, and flexibility -- make it a promising tool for privacy-sensitive domains such as healthcare and finance.

A new type of federated clustering: A non-model-sharing approach

TL;DR

DC-Clustering presents a privacy-preserving, non-model-sharing federated clustering framework capable of handling complex horizontal-vertical data partitions. It uses an anchor-guided construction of intermediate representations and a collaborative representation to perform either k-means or spectral clustering in a single round of communication with a central analyst. Empirical results on synthetic and open datasets show clustering performance close to centralized pooling, addressing a key gap in federated clustering for mixed data distributions. The approach offers practical value for privacy-sensitive domains by enabling knowledge discovery from distributed heterogeneous data while reducing communication overhead.

Abstract

In recent years, the growing need to leverage sensitive data across institutions has led to increased attention on federated learning (FL), a decentralized machine learning paradigm that enables model training without sharing raw data. However, existing FL-based clustering methods, known as federated clustering, typically assume simple data partitioning scenarios such as horizontal or vertical splits, and cannot handle more complex distributed structures. This study proposes data collaboration clustering (DC-Clustering), a novel federated clustering method that supports clustering over complex data partitioning scenarios where horizontal and vertical splits coexist. In DC-Clustering, each institution shares only intermediate representations instead of raw data, ensuring privacy preservation while enabling collaborative clustering. The method allows flexible selection between k-means and spectral clustering, and achieves final results with a single round of communication with the central server. We conducted extensive experiments using synthetic and open benchmark datasets. The results show that our method achieves clustering performance comparable to centralized clustering where all data are pooled. DC-Clustering addresses an important gap in current FL research by enabling effective knowledge discovery from distributed heterogeneous data. Its practical properties -- privacy preservation, communication efficiency, and flexibility -- make it a promising tool for privacy-sensitive domains such as healthcare and finance.

Paper Structure

This paper contains 28 sections, 2 theorems, 25 equations, 3 figures, 6 tables, 3 algorithms.

Table of Contents

  1. Introduction
  2. Related works
  3. Clustering
  4. Data Collaboration analysis
  5. Federated Clustering
  6. Preliminaries
  7. Clustering Algorithm
  8. k-means Clustering
  9. Spectral Clustering (SC)
  10. Distributed Data Setting
  11. Method
  12. Data Collaboration Clustering Framework
  13. Construction of collaborative representations
  14. Clustering and result sharing
  15. Discussion on Privacy-preserving of DC-Clustering
  16. ...and 13 more sections

Key Result

Theorem 1

Suppose that $f_i=(F_i,\mu_i )$ for $i=1,2$ satisfies the following conditions: where $\mathcal{R}$ denotes the range. Then, there exists a pair of affine transformations $(g_1,g_2 )$ that satisfies (eq:g_equal).

Figures (3)

  • Figure 1: Example of data distribution.
  • Figure 2: Overview of proposed method.
  • Figure 3: Distribution of generated synthetic data. (a) and (b) represent IID and non-IID setting, respectively, for Blobs, and (c) and (d) represents those for Circles. Orange and blue markers represent datasets with first-row parties and second-row parties, respectively.

Theorems & Definitions (4)

  • Theorem 1
  • proof
  • Corollary 1
  • proof