Work

An Innovative Reliable Client-Centric Deep Learning Inference Methodology

Year: 2024
Type: article
Abstract: Mobile phones and tablets have access to a very huge amount data that may be utilized to train learning models, potentially improving the user experience significantly. Nevertheless, the data available is often both extensive and sensitive, making it challenging to collect at centralize server and train within a centralized server using conventional methods. In this study, we investigate the utilization of blockchain technology with decentralized digital ledger to create a decentralized client-centric distributed learning system with the flexibility to support various machine learning models. This system enables the training of machine learning models directly on local machines, thereby addressing the constraints imposed by centralized servers. We demonstrate our system design, which includes two decentralized blockchain models built using Python Tensor Flow to ensure the system's reliability and efficiency. Ultimately, Block-CCL serves as an experimental environment for evaluating and distinguishing the impact of decentralized client centric i.e. federated learning from synchronization of model methods on the performance of the entire system. This highlights the validity and effectiveness of a federated learning system as a viable alternative to more centralized machine learning models.
Language: English
Open Access status: diamond
OpenAlex ID: W7217397018
FWCI:
Cites: 16
Cited by:
Related to: 10