I am Jing Zhu, a second year PhD student at the University of Michigan. My advisor is Danai Koutra. I have also interned at LLNL with Mark Heimann in summer 2021 and AWS DGL Team with Xiang Song , Vassilis N. Ioannidis and Christos Faloutsos in summer 2022. In 2023, I went back to LLNL to work on the exciting knowledge-infused gene prediction projects with Jay Thiagarajan , Mark Heimann and Christine Klymko. During my undergrad, I worked on multimodal learning with Andrew Owens, which inspired me a lot on my PhD research.
My past research includes unsupervised graph learning (SDM 2021, CIKM 2022), and multimodal learning (EMNLP 2021, NeurIPS 2022). Currently I work more on link prediction using a variety of methods (GNNs, KGEs, LMs). My goal is to build multimodal, structure-aware recommendation systems.
- [2022/09] "Touch and Go: Learning from Human-Collected Vision and Touch is accepted at NeurIPS'22!
- [2022/08] "CAPER: Coarsen, Align, Project, Refine - A General Multilevel Framework for Network Alignment is accepted at CIKM'22!
- [2021/08] "NegatER: Unsupervised Discovery of Negatives in Commonsense Knowledge Bases is accepted at EMNLP'21!
- [2020/12] "Node Proximity Is All You Need: A Unified Framework for Proximity-Preserving and Structural Node and Graph Embedding." is accepted at SDM'21!
- We present TouchUp-G; a simple Touch-Up enhancement technique to improve Graphs’ node features obtained from Pretrained Models(PMs) via graph-centric pretraining. TouchUp-G is especially helpful for recommendation tasks with various types of natural features.
- We propose SpotTarget: the first theoretical and empirical framework that analyzes on the effect of including target edges as message passing edges at training and test time.
- We propose Touch and Go: a multimodal visuo-tactile learning dataset which tries to associate sight with touch to understand material properties
- We propose CAPER, a multilevel alignment framework that improves upon existing alignment algorithms by enforcing alignment consistency across multiple graph resolutions.
- We propose NegatER, the first Unsupervised framework that ranks potential negatives in commonsense KBs using a contextual language model (LM).
- We present the first unified framework of node embedding (PhUSION) that includes both proximity-preserving and structural embeddings.
TouchUp-G: Improving Feature Representation through Graph-Centric Finetuning. [link]
SpotTarget: Rethinking the effect of target edges for link prediction in Graph Neural Networks.[link]
Touch and Go: Learning from Human-Collected Vision and Touch. [link]
CAPER: Coarsen, Align, Project, Refine - A General Multilevel Framework for Network Alignment. [link]
NegatER: Unsupervised Discovery of Negatives in Commonsense Knowledge Bases. [link]
Node Proximity Is All You Need: A Unified Framework for Proximity-Preserving and Structural Node and Graph Embedding. [link]
- CIKM Travel Award, SIGIR. Oct. 2022
- KDD Travel Award, SIGKDD. Aug. 2022
- CSE fellowship, University of Michigan. 2021-2022
- Rackham Travel Grant, University of Michigan. Nov. 2021.
- SDM Travel Award, SIAM. Apr. 2021.
- James B. Angell Scholar, University of Michigan. Mar. 2021.
- Jackson and Muriel Lum Scholarship, University of Michigan. 2019-2021
- Excellent Undergraduate Scholarship, Shanghai Jiao Tong University. Nov. 2018, Nov. 2019.
- EECS476 Data Mining, University of Michigan. 2021 Winter.
- EECS496 Major Design Experience-Professionalism, University of Michigan. 2020 Fall.
- VV285 Honors Mathematics III, Shanghai Jiao Tong University. 2019 Summer.
- VV186 Honors Mathematics II, Shanghai Jiao Tong University. 2018 Fall.
- Review: TKDD, NeurIPS, AAAI, KDD, ECML-PKDD
- I would love to help minority students in CS research. Feel free to reach out if you need anything.
- My research journey starts during high school with my great advisor SHI Guochao. We investigate the effect of Cactus Optunia on Bronchitis and won multiple innovation awards including Honorable Mention in Shing-Tung Yau High School Science Award.