Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Communication-Efficient Federated Learning Based on Explanation-Guided Pruning for Remote Sensing Image Classification

Jonas Klotz, Barış Büyüktaş, Begüm Demir

TL;DR

This work addresses communication bottlenecks in federated learning for remote sensing image classification by introducing an explanation-guided pruning method based on layer-wise relevance propagation (LRP). The central server prunes the least informative global model components using a learned relevance mask, applied after a warmup phase and reused across rounds to cut down transmitted updates. On BigEarthNet-S2 with FedAvg and a ResNet-50 backbone, the approach yields higher mAP compared to standard FL and random pruning across several pruning rates, while incurring minimal computational overhead. The method demonstrates robust gains in generalization and communication efficiency, and the authors outline future work on incorporating client importance into aggregation for further improvements.

Abstract

Federated learning (FL) is a decentralized machine learning paradigm in which multiple clients collaboratively train a global model by exchanging only model updates with the central server without sharing the local data of the clients. Due to the large volume of model updates required to be transmitted between clients and the central server, most FL systems are associated with high transfer costs (i.e., communication overhead). This issue is more critical for operational applications in remote sensing (RS), especially when large-scale RS data is processed and analyzed through FL systems with restricted communication bandwidth. To address this issue, we introduce an explanation-guided pruning strategy for communication-efficient FL in the context of RS image classification. Our pruning strategy is defined based on the layer-wise relevance propagation (LRP) driven explanations to: 1) efficiently and effectively identify the most relevant and informative model parameters (to be exchanged between clients and the central server); and 2) eliminate the non-informative ones to minimize the volume of model updates. The experimental results on the BigEarthNet-S2 dataset demonstrate that our strategy effectively reduces the number of shared model updates, while increasing the generalization ability of the global model. The code of this work is publicly available at https://git.tu-berlin.de/rsim/FL-LRP.

Communication-Efficient Federated Learning Based on Explanation-Guided Pruning for Remote Sensing Image Classification

TL;DR

This work addresses communication bottlenecks in federated learning for remote sensing image classification by introducing an explanation-guided pruning method based on layer-wise relevance propagation (LRP). The central server prunes the least informative global model components using a learned relevance mask, applied after a warmup phase and reused across rounds to cut down transmitted updates. On BigEarthNet-S2 with FedAvg and a ResNet-50 backbone, the approach yields higher mAP compared to standard FL and random pruning across several pruning rates, while incurring minimal computational overhead. The method demonstrates robust gains in generalization and communication efficiency, and the authors outline future work on incorporating client importance into aggregation for further improvements.

Abstract

Federated learning (FL) is a decentralized machine learning paradigm in which multiple clients collaboratively train a global model by exchanging only model updates with the central server without sharing the local data of the clients. Due to the large volume of model updates required to be transmitted between clients and the central server, most FL systems are associated with high transfer costs (i.e., communication overhead). This issue is more critical for operational applications in remote sensing (RS), especially when large-scale RS data is processed and analyzed through FL systems with restricted communication bandwidth. To address this issue, we introduce an explanation-guided pruning strategy for communication-efficient FL in the context of RS image classification. Our pruning strategy is defined based on the layer-wise relevance propagation (LRP) driven explanations to: 1) efficiently and effectively identify the most relevant and informative model parameters (to be exchanged between clients and the central server); and 2) eliminate the non-informative ones to minimize the volume of model updates. The experimental results on the BigEarthNet-S2 dataset demonstrate that our strategy effectively reduces the number of shared model updates, while increasing the generalization ability of the global model. The code of this work is publicly available at https://git.tu-berlin.de/rsim/FL-LRP.
Paper Structure (4 sections, 7 equations, 2 figures, 1 table)

This paper contains 4 sections, 7 equations, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Proposed Explanation-Guided Pruning Strategy for FL
  3. Experimental Results
  4. Conclusion

Figures (2)

  • Figure 1: An illustration of the pruning process and distribution of the pruned model to clients in the proposed strategy.
  • Figure 2: mAP scores (%) versus communication rounds obtained by the standard and the proposed strategies, considering different pruning rates.