Paper proposes one-shot federated density peak clustering with a shuffler
Original title: Evolving Federated Density Peak Clustering with Shuffler-Based Anonymization
Clients send only selected density peaks through a shuffler, and the method beat recent one-shot baselines on benchmarks.
The paper introduces Evolving Federated Density Peak Clustering (eFDPC), a one-shot federated clustering method that combines evolving fuzzy data-stream clustering with density peak clustering. Each client builds micro-clusters locally and transmits only selected peaks rather than full local models. A shuffler anonymizes client contributions before the server reconstructs them and selects global peaks. Variants are peak-only eFDPC-P and eFDPC-G with server-side label propagation.
- The design uses a single communication round to limit what the server sees.
- Density peak clustering runs on learned micro-clusters, not full local models.
- A shuffler sits between clients and the server to unlink contributions.
- eFDPC-G adds server-side label propagation; eFDPC-P is peak-only.
- Experiments use synthetic and real-world benchmarks, including non-IID data.
From the original article
1 min read at arxiv.orgIn this paper, we present a novel one-shot federated clustering approach based on evolving fuzzy clustering and shuffler-based anonymization. In federated clustering, each round reveals additional model information to the server and can increase the opportunity to infer properties of local datasets, even when raw samples are never shared.
By Miha Ožbot · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.