Project Description
This project aims to explore personalized federated learning (PFL) to enable adaptive AI models that cater to individual users while preserving data privacy. As an undergraduate research project, the objectives focus on developing a testbed for PFL to conduct preliminary experiments and facilitate future research on PFL. The project aims to set up a distributed computing environment using the open-source simulator to allocate virtual clients and a cloud server. By designing an efficient AI model training pipeline, we plan to implement a scalable framework to support model aggregation, personalization, and adaptation across heterogeneous clients. Finally, lightweight personalization techniques will be investigated that balance model performance and resource constraints on IoT/embedded systems. To achieve these objectives, this project will complete the testbed setup by deploying a federated learning framework using FLOWER to establish communication protocols for decentralized training and model aggregation. Baseline federated learning algorithms (e.g., FedAvg, FedProx) will be implemented and integrated with personalized learning approaches (e.g., local fine-tuning, meta-learning). Experiments will be conducted across heterogeneous clients with varying data distributions to measure model accuracy, convergence speed, and time consumption. This preliminary work will provide insights into the feasibility of deploying personalized AI models, laying the groundwork for larger-scale investigations.
Tasks and Responsibilites
The student's tasks and responsibilities include a literature review by conducting thorough research to understand existing methodologies and findings related to federated learning. This involves reading online tutorials, academic papers, articles, and other relevant materials to grasp the current state of the field. In addition, they are responsible for assisting in setting up experiments to measure power consumption during AI training sessions. This could involve working with specialized equipment for power measurement and ensuring data accuracy. Furthermore, the student will analyze the collected data using statistical tools and software, whose role might include organizing data, performing calculations, and generating visualizations to interpret the results effectively.