[Photo credit: Hui Shi]

Xiaofan Yu (于晓凡)

Assistant Professor, Department of Electrical Engineering
University of California, Merced

E-mail: xiaofanyu@ucmerced.edu
Office: Science and Engineering Building 2 (SE2) #382

Education:
Ph.D., UC San Diego, 2025, advised by Prof. Tajana Šimunić Rosing
B.S., Peking University, China, 2018

Awards:
ML&Systems Rising Star, 2024
CPS Rising Star, 2023
EECS Rising Star, 2022

More Information: My CV, Group's Website

Research Involvement in Centers and Grants

During my PhD, I was fortunate to work with the following research centers and funding sources:

JUMP 2.0 Research Centers (SRC & DARPA)

• CoCoSys – Center for the Co-Design of Cognitive Systems
• PRISM – Processing with Intelligent Storage and Memory

National AI Research Institutes (NSF)

• TILOS – The Institute for Learning-Enabled Optimization at Scale

NSF/Intel Partnerships

• MLWiNS – Machine Learning for Wireless Networking Systems

NSF Grants

• #1826967, #2100237, #2112167, #1911095, #2120019, #2211386

Research Collaborations

Through the support of these funding projects, I led several key collaborations that contributed to the majority of my PhD publications:

Robotics

• Prof. Vijay Kumar (University of Pennsylvania)
  

  Focus: Enabling adaptable and collaborative perception in multi-agent systems via federated learning.
  Approach: Federatedly learned models on multiple drones for predicting future trajectories of all targets from camera images and planned the future trajectories of each drone.
  Funding: TILOS
  Publications: arXiv'25

Natural Language Processing

• Prof. Larry Heck (Georgia Institute of Technology)
  

  Focus: Enabling natural language interactions between humans and sensors.
  Contribution: Created the first comprehensive dataset for sensor question answering and developed a pioneering real-time interaction system handling practical questions for long-term sensor data.
  Funding: CoCoSys
  Publications: IMWUT'25, SenSys'25, SenSys'24

Machine Learning

• Prof. Arya Mazumdar (UC San Diego)
  

  Focus: Optimizing Federated Learning in IoT networks under data and network heterogeneities.
  Approach: Designed theoretically sound and practical system solutions, tested on a real-world testbed with 40 clients built on Raspberry Pis and CPU clusters.
  Funding: TILOS
  Publications: IoTDI'23

Computer Vision

• Prof. Yunhui Guo (University of Texas at Dallas)
  

  Focus: Designing lifelong (continual) learning algorithms for edge devices adapting to real-world distribution drifts.
  Achievement: Developed new unsupervised lifelong learning algorithms for class-incremental streams, achieving up to 53.7% linear accuracy improvements over state-of-the-art methods.
  Publications: WACV'24, CVPRW'23, MSN'20

Industry Collaborations

• Intel Labs
  

  Focus: Designing Hyperdimensional Computing (HDC) algorithms for real-world challenges and applications.
  Achievement: Achieved similar or better accuracy compared to state-of-the-art neural network baselines while improving energy efficiency by up to 34.3×.
  Funding: MLWiNS
  Publications: IPSN'24, DATE'24, ASP-DAC'24, MobiCom'22

• Arm Research
  

  Focus: Optimizing long-term reliability of IoT networks via sensor deployment and network optimization.
  Achievement: Developed strategies that save up to 40% of maintenance costs compared to existing deployment heuristics while maintaining the same sensing quality.
  Publications: IoTDI'23, TNSM'22, CNSM'21, CNSM'20, TCAD'20, ICIOT'20