Portrait
Zhicheng Xu
Ph.D. Candidate
The University of Hong Kong
About Me

I am a Ph.D. candidate in the Department of Electrical and Computer Engineering at The University of Hong Kong (HKU), advised by Prof. Can Li.

My research focuses on in-memory computing and efficient hardware for AI, including compute-in-memory architectures based on emerging non-volatile memories (e.g., ferroelectric field-effect transistors and memristors), reliable analogue computing with imperfect devices, and hardware acceleration for Bayesian inference and nearest-neighbor search.

My work has been published in Nature Electronics, Nature Communications, IEEE Transactions on Computers, DAC, DATE, and ICCAD.

Education
  • The University of Hong Kong
    The University of Hong Kong
    Aug. 2023 – present
    Department of Electrical and Computer Engineering
    Ph.D. Candidate
  • Zhejiang University
    Zhejiang University
    Sep. 2019 – Jun. 2023
    B.S. in Microelectronics
News
2026
Our first-author paper is accepted to IEEE TCAD and will be presented at ESWEEK 2026 (CODES+ISSS) in Barcelona, Spain — see you in Barcelona this October!
Sep 07
Our paper Fault-free analogue computing with imperfect hardware is published in Nature Electronics! Read more
Jun 12
Our paper on trustworthy tree-based machine learning with MoS2 flash-based analog CAM is published in Nature Communications! Read more
Apr 24
2025
Our paper High-Performance In-Memory Bayesian Inference With Multi-Bit Ferroelectric FET is published in IEEE Transactions on Computers. Read more
Sep 15
2024
ShiftCAM, our time-domain content addressable memory for robust genome analysis, is presented at ICCAD 2024. Read more
Oct 27
FeBiM, our Bayesian inference engine with ferroelectric in-memory computing, is presented at DAC 2024. Read more
Jun 23
FeReX, our reconfigurable multi-bit ferroelectric compute-in-memory design for nearest neighbor search, is presented at DATE 2024. Read more
Mar 25
Selected Publications (view all )
Fault-free analogue computing with imperfect hardware
Fault-free analogue computing with imperfect hardware

Zhicheng Xu, Jiawei Liu, Sitao Huang, Zefan Li, Shengbo Wang, Bo Wen, Ruibin Mao, Mingrui Jiang, Giacomo Pedretti, Jim Ignowski, Kaibin Huang, Can Li

Nature Electronics 2026

Analogue in-memory computing computes directly within memory, but device failures and variations limit its precision and reliability. We report a fault-free matrix representation that decomposes any target matrix into two adaptive submatrices programmed onto the analogue hardware, allowing optimization to bypass faulty devices. Our memristor-based system achieves a cosine similarity of over 99.999% for a discrete Fourier transform matrix despite a 39% device fault rate.

[DOI]

Fault-free analogue computing with imperfect hardware

Zhicheng Xu, Jiawei Liu, Sitao Huang, Zefan Li, Shengbo Wang, Bo Wen, Ruibin Mao, Mingrui Jiang, Giacomo Pedretti, Jim Ignowski, Kaibin Huang, Can Li

Nature Electronics 2026

Analogue in-memory computing computes directly within memory, but device failures and variations limit its precision and reliability. We report a fault-free matrix representation that decomposes any target matrix into two adaptive submatrices programmed onto the analogue hardware, allowing optimization to bypass faulty devices. Our memristor-based system achieves a cosine similarity of over 99.999% for a discrete Fourier transform matrix despite a 39% device fault rate.

[DOI]

Trustworthy tree-based machine learning by MoS2 flash-based analog content-addressable memory with inherent soft boundaries
Trustworthy tree-based machine learning by MoS2 flash-based analog content-addressable memory with inherent soft boundaries

Bo Wen, Guoyun Gao, Zhicheng Xu, Mingrui Jiang, Ruibin Mao, Xiaojuan Qi, Jiezhi Chen, Xunzhao Yin, X. Sharon Hu, Can Li

Nature Communications 2026

Tree-based models such as Random Forest excel in interpretability and accuracy for tabular data, but scaling them on hardware is hard because sharp decision boundaries are highly susceptible to device variations. We present a hardware-software co-design using MoS2 flash-based analog content-addressable memory with inherent soft boundaries, enabling efficient and trustworthy inference with soft tree-based models.

[DOI]

Trustworthy tree-based machine learning by MoS2 flash-based analog content-addressable memory with inherent soft boundaries

Bo Wen, Guoyun Gao, Zhicheng Xu, Mingrui Jiang, Ruibin Mao, Xiaojuan Qi, Jiezhi Chen, Xunzhao Yin, X. Sharon Hu, Can Li

Nature Communications 2026

Tree-based models such as Random Forest excel in interpretability and accuracy for tabular data, but scaling them on hardware is hard because sharp decision boundaries are highly susceptible to device variations. We present a hardware-software co-design using MoS2 flash-based analog content-addressable memory with inherent soft boundaries, enabling efficient and trustworthy inference with soft tree-based models.

[DOI]

FeReX: A Reconfigurable Design of Multi-Bit Ferroelectric Compute-in-Memory for Nearest Neighbor Search
FeReX: A Reconfigurable Design of Multi-Bit Ferroelectric Compute-in-Memory for Nearest Neighbor Search

Zhicheng Xu, Che-Kai Liu, Chao Li, Ruibin Mao, Jianyi Yang, Thomas Kämpfe, Mohsen Imani, Can Li, Cheng Zhuo, Xunzhao Yin

Design, Automation & Test in Europe Conference (DATE) 2024

A reconfigurable multi-bit ferroelectric compute-in-memory design for nearest neighbor search.

[DOI]

FeReX: A Reconfigurable Design of Multi-Bit Ferroelectric Compute-in-Memory for Nearest Neighbor Search

Zhicheng Xu, Che-Kai Liu, Chao Li, Ruibin Mao, Jianyi Yang, Thomas Kämpfe, Mohsen Imani, Can Li, Cheng Zhuo, Xunzhao Yin

Design, Automation & Test in Europe Conference (DATE) 2024

A reconfigurable multi-bit ferroelectric compute-in-memory design for nearest neighbor search.

[DOI]

All publications