Direct Edge-to-Edge Cross-domain Model-based Transfer Learning

As technology in consumer goods production matures, the Internet of Things (IoT) systems have experienced rapid development, resulting in a massive amount of data that can be processed using deep neural networks. However, annotating the data and training the models require significant manpower, time, and computational resources. To address these issues, transfer learning can be utilized. Traditional systems rely on centralized servers for transfer learning. While our previous studies propose distributed systems for direct edge-to-edge transfer learning, they still neglect cross-domain model-based transfer learning, leading to lower learning efficiency and higher transmission costs. Therefore, this study introduces our plan to realize direct edge-to-edge model-based transfer learning across domains for edge devices. Preliminary experimental evaluations have demonstrated the effectiveness of our proposed approaches.