Abstract:In the realm of Federated Learning (FL) applied to remote sensing image classification, this study introduces and assesses several innovative communication strategies. Our exploration includes feature-centric communication, pseudo-weight amalgamation, and a combined method utilizing both weights and features. Experiments conducted on two public scene classification datasets unveil the effectiveness of these strategies, showcasing accelerated convergence, heightened privacy, and reduced network information exchange. This research provides valuable insights into the implications of feature-centric communication in FL, offering potential applications tailored for remote sensing scenarios.

What problem does this paper attempt to address?

The paper aims to address the application of Federated Learning (FL) in remote sensing image classification. Specifically, the researchers attempt to optimize FL's performance in remote sensing image classification tasks by introducing several innovative communication strategies. These strategies include feature-centric communication, pseudo-weight fusion, and a combination of weights and features. The main goal of the paper is to achieve accelerated convergence, enhanced privacy protection, and reduced network information exchange through these strategies, thereby providing potential application value for remote sensing scenarios. The background of the paper is that remote sensing applications typically require handling large amounts of data distributed across different geographical locations. Centralizing this data to train a centralized model may violate privacy regulations, while training models locally can lead to poor performance due to unevenly distributed data and overfitting issues. As the demand for complex image classification models grows, finding efficient and privacy-preserving methods becomes crucial. Federated Learning, as a potential solution, allows decentralized devices to collaboratively train machine learning models while keeping the original data localized, thus meeting this need. However, in remote sensing tasks, devices are often located far from the ground, limiting the amount of information that can be communicated. Therefore, FL methods suitable for small information volumes need to be developed. In this study, the authors explore various FL communication strategies based on the well-known FedA VG algorithm and investigate the trade-offs between reducing data transmission and performance. In particular, they propose sending and receiving only the average feature vectors for each category to minimize information exchange. Additionally, the study explores how combining feature communication and model parameter communication can achieve better performance and faster convergence. Experimental results show that these new strategies demonstrate effectiveness on two public scene classification datasets, proving their advantages in accelerating convergence, enhancing privacy protection, and reducing network information exchange.

Leveraging feature communication in federated learning for remote sensing image classification

Message Passing Based Wireless Federated Learning Via Analog Message Aggregation

Feature-based Federated Transfer Learning: Communication Efficiency, Robustness and Privacy

An Efficient Federated Learning Framework for Training Semantic Communication System

Exploring the Practicality of Federated Learning: A Survey Towards the Communication Perspective

Asynchronous Federated Learning over Wireless Communication Networks.

Training A Semantic Communication System with Federated Learning

Dynamic Sampling and Selective Masking for Communication-Efficient Federated Learning

Federated Local Compact Representation Communication: Framework and Application

Federated Learning Powered Semantic Communication for UAV Swarm Cooperation

Learning Across Decentralized Multi-Modal Remote Sensing Archives with Federated Learning

Computation and Communication Efficient Lightweighting Vertical Federated Learning

FLea: Addressing Data Scarcity and Label Skew in Federated Learning via Privacy-preserving Feature Augmentation

Towards Efficient Communications in Federated Learning: A Contemporary Survey

Communication-Efficient Federated Learning in Channel Constrained Internet of Things

Federated learning compression designed for lightweight communications

Federated Learning Across Decentralized and Unshared Archives for Remote Sensing Image Classification

Federated Learning: Strategies for Improving Communication Efficiency

Bridging Data Islands: Geographic Heterogeneity-Aware Federated Learning for Collaborative Remote Sensing Semantic Segmentation

Federated Learning Approach for Remote Sensing Scene Classification

Communication-Efficient Framework for Distributed Image Semantic Wireless Transmission