Yijiang (William) Li ¹ @ JHU

Looking for Ph.D. position Fall 2024!!

I am a Master’s student (M.S.E. in Data Science and Computer Science) at Johns Hopkins University(JHU). Prior to that, I received my Bachelor’s Degree in Computer Science at South China University of Technology, advised by Prof. Ying Gao.

I am currently a research assistant at Computational Cognition, Vision, and Learning (CCVL) research group, advised by Prof. Alan Yuille at JHU. I also work closely with Prof. Haohan Wang from UIUC at DREAM Lab. Prior to that, I spent one wonderful year of internship as a researcher in SenseTime Research and two months at Webank as a Machine Learning Engineer.

My research interest lies in sample efficient, trustworthy and generalizable Machine Learning. I am also interested in Computer Vision, intrigued by various aspects such as scene understanding, 3D vision models and cogsci-inspired vision. I am currently fascinated by the following research questions:

Can we use language to scaffold vision learning? The robust vision system humans possess can be partially credited to the human language which inherently captures high-level semantics and abstractions facilitating easy compositional reasoning.
How to better leverage 3D aware vision model to reason more robustly. Despite most vision models relying on 2D-pixel input, the world is intrinsically 3D. Beyond the sophisticated 3-D sensors, akin to human binocular vision, the critical aspect lies in developing a 3D aware structure behind the vision system that can model the real 3D world.

I am also broadly interested in robust decision-making, LLM as agents and AI4SCI ².

Contact: yli556 [at] jhu [dot] edu

🔥 News

2024.3: Our papers (preprint) “Beyond Finite Data: Towards Data-free Out-of-distribution Generalization via Extrapolation”, “Dataset distillation via the wasserstein metric” and “Towards Adversarially Robust Dataset Distillation by Curvature Regularization” is out on arxiv!
2023.12: Official code for Diverse Co-training is out on github!
2023.10: Our preprint paper “Understanding Adversarial Transferability in Federated Learning” is out on arxiv!
2023.10: 🎉 Two papers are presented at ICCV 2023, including one first-author paper.
2023.06: 🎉 Consistent-Teacher is presented at CVPR 2023 as Highlight (top 2.5%).
2023.02: I started my M.S.E. in Computer Science at Johns Hopkins University!
2022.06: 🎉 My first first-author paper is presented at BIBM 2022!

📝 Publications and Preprints

My full paper list can be seen at Google Scholar and Full List of Publications and Preprints.

Preprint

Y. Li, Y. Gao, and H. Wang, “Understanding adversarial transferability in federated learning,” arXiv preprint arXiv:2310.00616, 2023.

Preprint (TMLR submission)

Paper | Arxiv | BibTex

Preprint

Li Y, Ren S, Deng W, et al. Beyond Finite Data: Towards Data-free Out-of-distribution Generalization via Extrapolation. arXiv preprint arXiv:2403.05523.

Preprint (ECCV2024 submission)

Paper | Arxiv | BibTex

Preprint

E. Xue, Y. Li, H. Liu, Y. Shen, and H. Wang, “Towards adversarially robust dataset distillation by curvature regularization,” arXiv preprint arXiv: 2403.10045, 2024.

Preprint (ECCV2024 submission)

Paper | Arxiv | BibTex

Preprint

H. Liu, L. Li, T. Xing, V. Dalal, J. He, and H. Wang, “Dataset distillation via the wasserstein metric,” arXiv preprint arXiv: 2311.18531, 2023.

Preprint (ECCV2024 submission)

Paper | Arxiv | BibTex

ICCV 2023

Y. Li, X. Wang, L. Yang, L. Feng, W. Zhang, and Y. Gao, “Diverse co-training makes strong semi-supervised segmentor,” in Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pp. 16055–16067, 2023.

ICCV 2023

Paper | Code | Arxiv | BibTex

ICCV 2023

S. Huang, Y. Li, C. Chen, L. Shi, and Y. Gao, “Multi-metrics adaptively identifies backdoors in federated learning,” in Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pp. 4652–4662, October 2023.

ICCV 2023

Paper | Code | Arxiv | BibTex

CVPR 2023

X. Wang, X. Yang, S. Zhang, Y. Y. Li. Feng, S. Fang, C. Lyu, K. Chen, and W. Zhang, “Consistent-teacher: Towards reducing inconsis- tent pseudo-targets in semi-supervised object detection,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 3240–3249, 2023.

CVPR 2023 (Highlight Top 2.5%)

Project Page | Paper | Code | Arxiv | BibTex

BIBM 2022

Y. Li, W. Cai, Y. Gao, C. Li, and X. Hu, “More than encoder: Introducing transformer decoder to upsample,” in 2022 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pp. 1597–1602, IEEE, 2022.

BIBM 2022

Paper | Arxiv | BibTex

🎖 Honors and Awards

Outstanding Thesis Awards, awarded by South China University of Technology
2018-19 Scholarship, awarded by South China University of Technology
The 2020 Contemporary Undergraduate Mathematical Contest in Modeling First Prize
The 3rd Zhong Qing Bei National Mathematical Contest in Modeling ( undergraduate group) Second Prize
The 2020 May Day Mathematical Contest in Modeling First Prize

💻 Internships

Radiology and Radiological Science, Johns Hopkins Medicine | Research Assistant | May. 2023 – Sep. 2023

Synthesizing high-resolution vessel Wall MRI for visualization of Lenticulostriate Arteries
Implement training and inference pipeline for PD to WIP modality conversion.
Implement UNet, CGAN, Cycle GAN as baseline methods.
Propose and implement condiction DDPM for PD to WIP modality conversion.

Sensetime Research | Researcher | Jul. 2021 – Nov. 2022

Change detection in remote sensing reaching 80% in precision and recall.
Reproduce several SOTA papers in semantic segmentation and remote sensing.
Research: sample efficiency (self-supervised, semi-supervised and weakly supervised learning).

WeBank | Machine Learning Engineer | Jun. 2020 - Sep. 2020

Public Sentiment and Risk Control Project: text analysis based on big data, distributed training, and Hadoop clusters.
Train an NLP model on web-crawled news text based on bert, xgboost, etc. and build a textual processing system based on Hadoop clusters; conduct spam classification and sentiment analysis.
Iteratively improve the trained spam classification model with real-time data and problems.

📖 Educations

Johns Hopkins University | M.S.E. in Data Science
Jan. 2023 – Present

GPA: 4.0 / 4.0

South China University of Technology | B.E. in Computer Science
Sep. 2018 – Jul. 2022

GPA: 3.73 /4.0
2018-19 Scholarship
Outstanding Thesis Awards

The Affiliated High School of South China Normal University | High School Diploma
Sep. 2015 – Jul. 2018

📚 Service

Teaching

EN.553.436/636 Introduction to Data Science (Fall 2023)
- Teaching Assistant
- Instructor: Prof. Tamás Budavári
EN 601.482/682 Machine Learning: Deep Learning (Fall 2023)
- Course Assistant
- Instructor: Prof. Mathias Unberath
EN.553.436/636 Introduction to Data Science (Spring 2023)
- Teaching Assistant
- Instructor: Prof. Tamás Budavári, Prof. Soledad Villar