Haowei Li

I am a master's student in Communications Engineering at Xidian University, under the guidance of Professor Weiying Xie. I obtained my bachelor degree from Xidian University.

I am very interested in various fields of machine learning. Currently, my research focus is primarily on developing efficient algorithms for distributed machine learning.

My hobbies include singing pop music and working out.

Email  /  Github

Xidian University, Xi’an, China09/2023-Now

• Department: School of Telecommunications Engineering (M.E.)
• Major: Electronic and Information Engineering
• Advisor: Prof. Weiying Xie
• Outstanding Engineer Experimental Class

Xidian University, Xi’an, China09/2019-06/2023

• Department: School of Telecommunications Engineering (B.E.)
• Major: Communication Engineering
• Outstanding Engineer Experimental Class

• Distributed Machine Learning
• Federated Learning
• Large-scale Models
• Optimization algorithms

Author: Weiying Xie* (Advisor), Haowei Li*, Jitao Ma, Yunsong Li, Jie Lei, Donglai Liu and Leyuan Fang.

• We construct a Sparsification-Quantization joint learning framework to compress communication in distributed machine learning.
• Has been accepted in IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR- 2024).
• Code Repository: https://github.com/HaoweiLi778/JointSQ

Author: Haowei Li, Weiying Xie (Advisor), Hangyu Ye, Jitao Ma, Shuran Ma and Yunsong Li.

• We devise personalized compression strategies at the parameter level for each client to address the Non-IID characteristics of Federated Learning.
• Under review by THE THIRTY-NINTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE (AAAI-2025).
• Paper link: https://arxiv.org/abs/2408.08977

Author: Shuran Ma, Weiying Xie (Advisor), Daixun Li, Haowei Li, Yunsong Li.

• We propose FedCD, a framework tackling cross-domain generalization in federated learning with IMG and REA methods. FedCD outperforms baselines, improving accuracy by 1.45% and mAP50 by 4.36%.
• Paper link: https://arxiv.org/abs/2407.19174

• We are using SAM in the field of multimodal fusion, and we propose a cross-domain prompt learning model.
• We have designed a multi-satellite intelligent interpretation system for the distributed deployment of SAM.
• The relevant work has been implemented for coastline extraction in Uzbekistan. The average accuracy obtained from remote sensing data in Uzbekistan is 91.64%.

• We present the unified Mars visual language instruction dataset, Mars-VL-227k.
• We developed the Chat-Mars-LLAMA3 model, which is not only the first deep space remote sensing model based on LLAMA3 but also the first to unify Earth observation and deep space Mars visual language tasks.

• IEEE student member.
• Graduate Outstanding Academic Scholarship, Xidian University (2024).
• Undergraduate School-Level Scholarship, Xidian University (2022).
• Third Prize in the 2nd "Fire Cup" Target Intelligent Recognition Competition (2023).

• Fibocom Wireless Inc, Xi’an, China03/2022-07/2022
• Department: Software Development (Intern)
• Position: Project Leader of the “Facial Recognition Attendance System”
• We are responsible for establishing an employee facial database and developing a facial recognition system on a demo board, based on FaceNet (an open-source project utilizing CNN). Additionally, we are developing an APP to provide user APIs and handle the system's daily maintenance and updates.

• Here are the photos I took when I attended CVPR 2024:[here]
  
• This is the student club I joined, and it's about pop music:[here]