Comparison Summary of Centralized learning and Federated learning

Децентрализация обработки изображений путем интеграции федеративного обучения и сверточных нейронных сетей

Table 1 - Comparison Summary of Centralized learning and Federated learning

Aspect	Centralized Learning	Federated Learning
Data Movement	Data is transferred to a master server	Data is kept on local devices
Privacy	High probability of data leakage	Improved data security and privacy
Communication Overhead	High, particularly for image data	less, only model parameters are shared
Scalability	Bandwidth and server capacity limited	Highly scalable to distributed clients
Fault Tolerance	Single point of failure	Higher, with distributed nodes
Deployment	Easier in controlled systems	Realistic to real-world, decentralized systems