Accelerating AI Model Training with Collaborative Computing: A Survey

In this article, the authors explore the challenges of training artificial intelligence models in edge networks, where data is decentralized and distributed across multiple devices. They propose a novel approach called collaborative computing framework, which enables the training of accurate models while minimizing overall time consumption. The proposed framework leverages a game-theoretic approach to optimize resource allocation and incentive mechanisms, ensuring fairness and efficiency among participating devices.
The authors begin by highlighting the limitations of traditional centralized learning methods in edge networks, where data is scattered and computing resources are limited. They introduce the concept of federated learning, which enables multiple devices to collaboratively train a shared model without sharing raw data. However, this approach also faces challenges related to resource allocation and incentive mechanisms, as devices may have conflicting interests.
To address these issues, the authors propose a novel framework that leverages game theory to optimize resource allocation and design incentive mechanisms. They demonstrate through simulations that their proposed framework can achieve an accuracy of over 95% while minimizing overall time consumption. The authors also analyze the system’s performance under different scenarios and show that their approach can handle varying network conditions and device capacities.
The article provides a comprehensive overview of existing studies on federated learning and identifies key challenges in edge networks. The proposed framework offers a promising solution to these challenges, providing a more efficient and effective way to train AI models in decentralized environments. Overall, the article demonstrates the potential of collaborative computing frameworks for enhancing the performance of AI systems in edge networks.