CV

Education


Carnegie Mellon University
PhD Candidate in Electrical & Computer Engineering
Thesis: Optimizing Decentralized Learning - System and Device-Level Strategies
University of North Carolina at Chapel Hill, 2019
B.S in Physics & B.S in Computer Science with Highest Honors
Minor in Mathematics
Honors Thesis: Unsupervised Learning of the Phase Transitions for 2D Ising Model
Honors Thesis: A Frameless Image Sensor with Application in Astronomy

Current Research


Ph.D. Research, Carnegie Mellon University
Department of Electrical & Computer Engineering
Advisors: Carlee Joe-Wong, Bob Iannucci
  • Employ techniques like reinforcement learning, graph neural networks, and multi-armed bandits to reduce computation and communication overhead.
  • Develop solutions to optimize energy usage at both the device and network levels, improving overall system performance and scalability in IoT/edge applications.
  • Investigate collaborative learning frameworks such as federated learning, addressing challenges of distributed learning tasks across networks of devices with heterogeneous data distribution and varying computational capabilities.

Publications


Y. Hu, B. Iannucci, and C. Joe-Wong. Towards Adaptive Optimal Placement in Heterogeneous Computing. (under review).

Y. Hu, I. Lin, J. Zuo, B. Iannucci, and C. Joe-Wong. QuBA-FL: Bitwidth-Adaptive Quantized Training for Efficient Federated Learning at the Edge. (submitted)

Y. Hu, J. Zuo, E. Zhang, B. Iannucci, and C. Joe-Wong. Tin-Tin: Towards Tiny Learning on Tiny Devices with Integer-based Neural Network Training. (submitted)

Y. Hu, J. Zuo, A. Zhao, B. Iannucci, and C. Joe-Wong (2024). CoRAST: Towards Foundation Model-Powered Correlated Data Analysis in Resource-Constrained CPS and IoT, 2024 International Workshop on Foundation Models for Cyber-Physical Systems & Internet of Things (FMSys), Hong Kong, 2024 pp. 1-6. doi: 10.1109/FMSys62467.2024.00005.

Y. Hu, J. Zuo, B. Iannucci, and C. Joe-Wong (2023). Intelligent Communication Planning for Constrained Environmental IoT Sensing with Reinforcement Learning, 2023 20th Annual IEEE International Conference on Sensing, Communication, and Networking (SECON), Madrid, Spain, 2023, pp. 357-365, doi: 10.1109/SECON58729.2023.10287493.

Y. Hu, C. Zhang, E. Andert, H. Singh, A. Shrivastava, J. Laudon, Y. Zhou, B. Iannucci, and C. Joe-Wong (2023). GiPH: Generalizable Placement Learning for Adaptive Heterogeneous Computing, Proceedings of Machine Learning and Systems (MLSys), 5.

M. Khayatian, M. Mehrabian, E. Andert, R. Grimsley, K. Liang, Y. Hu, I. Mccormack, C. Joe-Wong, J. Aldrich, B. Iannucci and A. Shrivastava (2022). Plan B - Design Methodology for Cyber-Physical Systems Robust to Timing Failures, Transactions on Cyber-Physical Systems (TCPS). doi: 10.1145/3516449

N. Rodríguez and Y. Hu (2020). On the steady-states of a two-species non-local cross-diffusion model, Journal of Applied Analysis, 26(1), 1-19. doi: 10.1515/jaa-2020-2003.

Awards & Fellowships


Past Experience


Data Science Intern, Ericsson Inc.
Ericsson Global AI Accelerator Group, 2021
Safe Reinforcement Learning for Network Optimization
  • Investigated the tradeoff between maximizing total network throughput and maintaining user quality of service (QoS) during remote electrical tilt adjustment in cellular networks.
  • Evaluated and implemented multiple safe RL algorithms in simulation environments to optimize antenna tilt dynamically while adhering to QoS requirements.
Student Researcher, University of North Carolina at Chapel Hill
Department of Physics, 2019. Advisor: Joaquín E Drut
Unsupervised Learning of Phase Transitions for Quantum Models (Thesis)
  • Applied unsupervised learning to identify phase transitions in lattice spin systems.
  • Used principal component analysis (PCA) and Monte Carlo simulations to clarify phase boundaries via dimensionality reduction.
Department of Computer Science, 2018. Advisor: Montek Singh
Asynchronous Frameless Imaging for Astronomy (Thesis)
  • Tackled the design challenge of creating a frameless image sensor for astronomy, aiming for high-resolution imaging in ultra-low-light environments, while minimizing sensitivity to noise.
  • Developed methods to lower noise floor and enabled single-photon detection in ultra-low-light environments.
Department of Mathematics, 2018. Advisor: Nancy Rodríguez
Cross-Diffusion Models of Species Behavior
  • Investigated species behavior in response to environmental constraints to model spatial dynamics and interactions affecting population distribution.
  • Simulated interspecies interactions in MATLAB using the finite element method (FEM), contributing to a journal publication.

Teaching Experience


Invited Lecture on Federated Learning, Duke University ECE Department, 2024
Course: Edge Computing
Teaching Assistant, CMU ECE Department, 2021
Course: Mathematical Foundations of Electrical Engineering
Teaching Assistant, CMU ECE Department, 2020
Course: How to Write Low-Power Code for the IoT?