Federated Learning: Collaborative Ml without ...

Federated Learning: Collaborative Ml without Centralized Data

Author Name : Venkata Naga Sai Kiran Challa

ABSTRACT Federated Learning (FL) is a decentralized approach to training machine learning models, allowing multiple parties to collaboratively train a shared model without sharing their data directly. This paper, authored by Venkata Naga Sai Kiran Challa, explores the advantages of FL, such as privacy preservation, data sovereignty, scalability, and reduced communication costs. It also discusses the challenges, including communication and synchronization, heterogeneity, and security concerns. The integration of blockchain technology with FL is proposed to enhance data integrity, auditability, and incentive mechanisms through decentralized consensus and smart contracts. An example scenario involving self-driving vehicles demonstrates the implementation of FL and blockchain, highlighting the process of local training, model aggregation, and incentivization. The paper compares FL with traditional centralized methods, underscoring the benefits and drawbacks of each approach