Scholarlink

<p>The rapid expansion of Internet of Things (IoT) applications has underscored the critical role of edge computing in enhancing real-time data processing and responsiveness. Although deploying advanced generative AI models such as GANs and VAEs at the edge holds immense potential, the inherent constraints in computational power, memory, and energy resources pose significant challenges. This work introduces a novel framework that integrates distributed machine learning techniques, including federated learning (FL), to optimize the deployment of generative AI models in resource-constrained edge environments. Importantly, FL is utilized here not for data privacy, but for its ability to improve efficiency and scalability by minimizing data transfers and leveraging decentralized data processing. The proposed framework incorporates strategies for power-aware orchestration, dynamic resource allocation, thermal management, and robust failure recovery. Case studies across various domains, such as healthcare, agriculture, and intelligent transportation systems, demonstrate the practical advantages of this approach, including energy savings and improved model performance. This work contributes to the advancement of sustainable AI-driven edge computing, highlighting the scalability and adaptability of our framework in diverse IoT settings.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Energy-efficient edge deployment of generative AI models using federated learning

  • 期刊论文

  • 发表日期：

    2025年4月28日

查看全文
  • Shadi Alzu’bi,
  • Tarek Kanan,
  • Mohammed Elbes,
  • Ghassan Kanaan,
  • Issam Trrad

摘要

The rapid expansion of Internet of Things (IoT) applications has underscored the critical role of edge computing in enhancing real-time data processing and responsiveness. Although deploying advanced generative AI models such as GANs and VAEs at the edge holds immense potential, the inherent constraints in computational power, memory, and energy resources pose significant challenges. This work introduces a novel framework that integrates distributed machine learning techniques, including federated learning (FL), to optimize the deployment of generative AI models in resource-constrained edge environments. Importantly, FL is utilized here not for data privacy, but for its ability to improve efficiency and scalability by minimizing data transfers and leveraging decentralized data processing. The proposed framework incorporates strategies for power-aware orchestration, dynamic resource allocation, thermal management, and robust failure recovery. Case studies across various domains, such as healthcare, agriculture, and intelligent transportation systems, demonstrate the practical advantages of this approach, including energy savings and improved model performance. This work contributes to the advancement of sustainable AI-driven edge computing, highlighting the scalability and adaptability of our framework in diverse IoT settings.