Local sharpness aware minimization in decentralized federated learning with privacy protection

Key Points

• Decentralized federated learning enhances privacy protection measures, and promotes data security during training.
• The model shows a significant reduction in training loss by 20% compared to traditional methods.
• Assessment using sharpness aware minimization in a federated setup improves overall model accuracy and robustness.
• These findings suggest a practical approach for implementing privacy-preserving mechanisms in AI systems.