About Me
Xiaoke Zhu · Research Fellow, Shenzhen Institute of Computing Sciences (SICS)
I am Xiaoke Zhu (Hsiaoko Chu in Wade-Giles romanization), currently a research fellow at Shenzhen Institute of Computing Sciences (SICS). I received my Ph.D. from Beihang University in October 2025, advised by Prof. Wenfei Fan, and my M.S. from Yunnan University in 2020. My work focuses on graph computing, GPU-accelerated algorithms, high-performance data cleaning, and AI4DB. I welcome collaborations and discussions—feel free to reach out via email.
News
- NOV 2025SIGMOD'26 Our paper Enumerating Graph Pattern Matches with ML Oracles was accepted to SIGMOD 2026.
- OCT 2025 I joined Shenzhen Institute of Computing Sciences (SICS) as a research fellow.
- NOV 2024 Successfully defended my Ph.D. thesis. (Download slides)
- NOV 2024SIGMOD'25 Our paper GPU-Accelerated Graph Cleaning with a Single Machine was accepted to SIGMOD 2025.
- NOV 2024VLDB'25 Our paper A Single Machine System for Querying Big Graphs with PRAM was accepted to VLDB 2025.
- OCT 2024VLDB'25 Our paper HyperBlocker: Accelerating Rule-Based Blocking in Entity Resolution Using GPUs was accepted to VLDB 2025.
- OCT 2024BigData'24 Our paper Deep Learning Service for Efficient Data Distribution Aware Sorting was accepted to IEEE BigData 2024.
- APR 2023VLDB'23 Our paper MiniGraph: Querying Big Graphs with a Single Machine was accepted to VLDB 2023.
- FEB 2022ICDE'22 Our paper Deep and Collective Entity Resolution in Parallel was accepted to ICDE 2022.
- SEP 2021 I joined Shenzhen Institute of Computing Sciences (SICS) as a research intern.
- JUL 2021IEEE CLOUD'21 Our paper DLB: Deep Learning Based Load Balancing was accepted to IEEE CLOUD 2021.
Research Interests
Graph Computing Systems
Building high-level programming models and runtime systems that execute graph applications on shared-memory, out-of-memory, CPU, or GPU architectures. I have improved I/O efficiency for out-of-core graph analytics (e.g., PageRank, SSSP) and optimized GPU performance for graph mining tasks such as graph cleaning and pattern matching.
Data Cleaning
Accelerating data cleaning systems on modern hardware such as GPUs and distributed clusters. I have benchmarked parallel runtime systems for data cleaning and identified their performance bottlenecks, with a focus on rule-based blocking and entity resolution.
AI4DB
Leveraging machine learning and deep learning to replace manual DBA effort and classical algorithms, enabling more efficient data processing and resource management. I have designed learned models for sorting, load balancing, and scheduling.
Representative work: IEEE CLOUD'21, IEEE BigData'24