Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Dynamic Clustering in Federated Learning

Yeongwoo Kim, Ezeddin Al Hakim, Johan Haraldson, Henrik Eriksson, José Mairton B. da Silva, Carlo Fischione

TL;DR

The generative adversarial network-based clustering preserves privacy, the cluster calibration deals with dynamic environments by modifying clusters, and the divisive clustering explores the different number of clusters by repeatedly selecting and dividing a cluster into multiple clusters.

Abstract

In the resource management of wireless networks, Federated Learning has been used to predict handovers. However, non-independent and identically distributed data degrade the accuracy performance of such predictions. To overcome the problem, Federated Learning can leverage data clustering algorithms and build a machine learning model for each cluster. However, traditional data clustering algorithms, when applied to the handover prediction, exhibit three main limitations: the risk of data privacy breach, the fixed shape of clusters, and the non-adaptive number of clusters. To overcome these limitations, in this paper, we propose a three-phased data clustering algorithm, namely: generative adversarial network-based clustering, cluster calibration, and cluster division. We show that the generative adversarial network-based clustering preserves privacy. The cluster calibration deals with dynamic environments by modifying clusters. Moreover, the divisive clustering explores the different number of clusters by repeatedly selecting and dividing a cluster into multiple clusters. A baseline algorithm and our algorithm are tested on a time series forecasting task. We show that our algorithm improves the performance of forecasting models, including cellular network handover, by 43%.

Dynamic Clustering in Federated Learning

TL;DR

The generative adversarial network-based clustering preserves privacy, the cluster calibration deals with dynamic environments by modifying clusters, and the divisive clustering explores the different number of clusters by repeatedly selecting and dividing a cluster into multiple clusters.

Abstract

In the resource management of wireless networks, Federated Learning has been used to predict handovers. However, non-independent and identically distributed data degrade the accuracy performance of such predictions. To overcome the problem, Federated Learning can leverage data clustering algorithms and build a machine learning model for each cluster. However, traditional data clustering algorithms, when applied to the handover prediction, exhibit three main limitations: the risk of data privacy breach, the fixed shape of clusters, and the non-adaptive number of clusters. To overcome these limitations, in this paper, we propose a three-phased data clustering algorithm, namely: generative adversarial network-based clustering, cluster calibration, and cluster division. We show that the generative adversarial network-based clustering preserves privacy. The cluster calibration deals with dynamic environments by modifying clusters. Moreover, the divisive clustering explores the different number of clusters by repeatedly selecting and dividing a cluster into multiple clusters. A baseline algorithm and our algorithm are tested on a time series forecasting task. We show that our algorithm improves the performance of forecasting models, including cellular network handover, by 43%.

Paper Structure

This paper contains 16 sections, 6 equations, 3 figures, 4 tables, 4 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Background
  4. Federated Learning
  5. ClusterGAN
  6. Hypothesis-based clustering (HypCluster)
  7. Divisive clustering
  8. Dynamic GAN-based Clustering
  9. Experimental Setting and Results
  10. Experimental Setup
  11. Evaluation metrics
  12. Results
  13. Result of Phase 1
  14. Result of Phase 2
  15. Result of Phase 3
  16. ...and 1 more sections

Figures (3)

  • Figure 1: The structure of the clusterGAN with the generator, encoder, and discriminator. The generator synthesizes data from a latent space, and the encoder maps the synthetic data into the latent space. The discriminator distinguishes between real data and synthetic data.
  • Figure 2: Dynamic GAN-based clustering consists of three phases and executes the three phases sequentially. When the cluster division selects a cluster to divide, our algorithm performs Phase 1 and 2 on the selected cluster.
  • Figure 3: LSTM training loss in the first divisive round of (a) Italy power demand, (b) Pendigit, (c) Melbourne pedestrian, and (d) Handover. The solid and dashed lines mean the cluster calibration in the static and dynamic environments.