Shiyu Zhao 赵时予

Graduate Student

Stanford University

Biography

I am currently a master student in computer science at Stanford University. I am experienced in solving complex problems in the real life through sound and creative engineering, and I am a quick learner who is always eager to explore new things! As for research, my previous research mainly lies in the field of knowledge discovery in AI, with a focus on the use of AI to effectively, efficiently, and interpretably gain knowledge in areas such as natural language processing, graph representation learning, and computational biology.

I am passionate about applying my computer science skills in the real life to make a difference and developing AI models that can uncover generalizable knowledge from data with complex structures and logic.

Interests

Natural language processing
Graph representation learning
Computational biology

Education

Stanford University, Sep 2023 - Jun 2025 (expected)
Computer Science (Master's degree) at School of Engineering
Tsinghua University, Sep 2019 - Jun 2023
Computer Science and Technology (Bachelor's degree) at Yao Class, Institute for Interdisciplinary Information Sciences
Yinchuan No.1 Middle School, Sep 2016 - Jun 2019

Experience

Software Engineer Intern

miHoYo

Jul 2023 – Aug 2032 Beijing, China

Mentored by Yinhe Zheng.

Remote Research Assistant

University of California San Diego

Jul 2022 – Nov 2022 San Diego, United States

Advised by Prof. Leon Bergen.

Machine Learning Intern

MILA-Quebec AI Institute

Mar 2022 – Aug 2022 Montreal, Canada

Mentored by Prof. Jian Tang.

Research Assistant

Tsinghua University

Jan 2021 – Dec 2021 Beijing, China

Advised by Prof. Jie Tang and Prof. Yuxiao Dong at Knowledge Engineering Group (KEG) and ZhipuAI

Publications

Xiao Liu, Shiyu Zhao, Kai Su, Yukuo Cen, Jiezhong Qiu, Mengdi Zhang, Wei Wu, Yuxiao Dong, Jie Tang (2022). Mask and Reason: Pre-Training Knowledge Graph Transformers for Complex Logical Queries. In KDD 2022.

PDF Cite Code Slides DOI arxiv

Research

LogicGNN: Logic Message Passing Graph Neural Network

The first systematically generalizable and scalable reasoning model on KG. It models one-step logic inference as triangle update on graph inspired by logic programming and formalize it under the designed GNN framework. By recursively applying triangle updates, the model could reason with logic rules of any length with strong scalability. It can also identify reasonable patterns and conduct partial reasoning with the help of auxiliary edges on graphs.

End-to-end Small Molecule Entity Discovery

A model-agnostic enhancement pipeline for small molecule property prediction. It uses a encoder and a retriever to retrieve similar molecules out of the whole dataset as supportive evidence. It also introduces a refinement stage to utilize only supportive evidence for downstream task. It will soon be sealed as a toolkit.

Knowledge Graph System: Graph Reasoning with GNN Matching

A graph-matching model with theoretical guarantees. It proposes that checking the condition of a rule is equivalent to performing subgraph isomorphism tests. It trains a graph neural network (GNN) on the line graph and exploited the trained model with matching-based grading. The extracted rules server as a tool for the interpretability analysis.

Mask and Reason: Pre-Training Knowledge Graph Transformers for Complex Logical Queries

A well-functioning pretrain-finetune paradigm on knowledge graph with great generalizability and interpretability. It formulates complex logical queries as masked predictions on graph patterns and introduces a two-stage masked pre-training strategy. It also proposes a KG triple transformation method to enable transformer to handle KG elegantly and a mechanism to unify different tasks of knowledge graph problems.

Selected Projects

A Survey on Non-black-box Simulator of Zero-knowledge Interactive proofs

Surveys over FLS-type protocal and its application over non-black-box simulator in zero-knowledge proof.

Comprehensive and distinguishable graph-linked embedding for multi-omics single-cell data integration

It solve the indistinguishability of aggregating multi-omics data on the graph for the graph-linked embedding. Besides, it enriches the multi-omics information of graph embedding by using multiple aggregators in the GNN