CV
This is my curriculum vitae. You can download it by clicking the top pdf download button.
Contact Information
| Name | Xiangyu Zhou |
| Professional Title | PhD student |
| xiangyuzhou823@gmail.com | |
| Location | 781 S Terrace Rd, Tempe, Arizona AZ 85281 |
| Website | https://shaun-z.github.io/ |
Professional Summary
I am a student at Arizona State University, majoring in Electrical Engineering. I am passionate about machine learning. I have a strong background in electrical engineering and computer science.
Experience
-
2023 - Research Assistant
Ira A. Fulton Schools of Engineering, Arizona State University
I am a research assistant at the Ira A. Fulton Schools of Engineering at Arizona State University.
- Research on Explainable Artificial Intelligence (XAI)
- Machine Learning
-
2022 - 2023 Teaching Assistant
Science & Technology Innovation Center, Huazhong University of Science and Technology
I was responsible for teaching students about electronic design and the usage of Xilinx Vivado development kit.
- Analyzed the topics and award-winning works in previous competitions and prepared tutorial materials in advance
- Collected needs of participants, contacted with suppliers of materials and made a training outline
- Prepared presentations and delivered trainings to new contestants in the pre-competition training process
- In charge of tutoring participants, answering their questions and offering technical support
- Be responsible for primary evaluation of deliveries, projects scoring and notes-making and modification of submitted works in the future three months
-
2022 - 2022 Intern Cloud Engineer
China Telecom Xiantao Branch
I was responsible for monitoring the network and maintaining the network equipment.
- Be responsible for troubleshooting and handling of network failures of customers and companies
- Assisted to conduct network construction and management, network architecture planning, design, adjustment and performance optimization
- Assisted Engineers to provide the following services to new customers
- Completed other temporary tasks assigned by leaders including preparing presentation materials, summarizing general network failures and corresponding solutions, documentations issues, etc.
-
2019 - 2020 Vice Minister
Academic Development Support Center, HUST
I helped professors and tutors complete teaching activities, conveyed students’ opinions to Department of Study; kept close contact with professors of all disciplines, and conveyed teachers’ notices and arrangements to the whole class
-
2020 - 2020 Interviewer
5 Hospitals in Wuhan
I interviewed doctors and nurses from 5 hospitals in Wuhan about the measures taken by the medical system in Wuhan during the early outbreak of Covid-19 and the feelings of frontline medical staff.
- Covid-19
- Medical System
- Interview
Education
-
2026 - Tempe, Arizona, US
Ph.D.
Arizona State University
Computer Engineering
- Explainable Artificial Intelligence
- Deep Learning
- Large Vision Language Models
-
2023 - 2025 Tempe, Arizona, US
M.S.
Arizona State University
Electrical Engineering
- Explainable Artificial Intelligence
- Deep Learning
- Computer Vision
- Electrical Engineering
-
2019 - 2023 Wuhan, China
B.S.
Huazhong University of Science and Technology
Telecommunications Engineering
- Calculus (90,92/100)
- Linear Algebra (94/100)
- Probability Theory and Mathematical Statistics (97/100)
- Physics (96,92/100)
- Analog Circuit and Digital System (95/100)
Awards
-
2022 National Second Prize
China College Students 'Internet+' Innovation and Entrepreneurship Competition Committee
A national level competition authorized by Ministry of Education.
-
2021 National Second Prize
National Undergraduate Electronic Design Competition Organization Committee
The competition is a national competition for undergraduate students in China, organized by the Ministry of Education and Ministry of Industry of China.
-
2020 -
2020 -
2019 Ranked 7th out of 136 teams
'Renesas' HUST Undergraduates Intelligent Vehicle Championship
Our team attended the ‘Advanced Group’, aiming at making an intelligent vehicle that can independently identify the path and finish the whole runway.
Publications
-
2025 Owen-based Semantics and Hierarchy-Aware Explanation (O-Shap)
2025 IEEE International Conference on Data Mining (ICDM)
Shapley value-based methods have become foundational in explainable artificial intelligence (XAI), offering theoretically grounded feature attributions through cooperative game theory. However, in practice, particularly in vision tasks, the assumption of feature independence breaks down, as features (i.e., pixels) often exhibit strong spatial and semantic dependencies. To address this, modern SHAP implementations now include the Owen value, a hierarchical generalization of the Shapley value that supports group attributions. While the Owen value preserves the foundations of Shapley values, its effectiveness critically depends on how feature groups are defined. We show that commonly used segmentations (e.g., axis-aligned or SLIC) violate key consistency properties, and propose a new segmentation approach that satisfies the T-property to ensure semantic alignment across hierarchy levels. This hierarchy enables computational pruning while improving attribution accuracy and interpretability. Experiments on image and tabular datasets demonstrate that O-Shap outperforms baseline SHAP variants in attribution precision, semantic coherence, and runtime efficiency, especially when structure matters.
Skills
Languages
Interests
Projects
-
LCCF: Layer-wise Concept Capture and Fusion for Interpretable Deep Learning
Create a lightweight and efficient framework for capturing and fusing concepts across different layers of deep neural networks, enabling enhanced interpretability without significant computational overhead.
- Design and implement novel algorithms for concept extraction and fusion, demonstrating their effectiveness on benchmark datasets (ImageNet) and real-world applications (medical imaging)
- Conduct extensive experiments to evaluate the performance of LCCF in terms of interpretability, accuracy, and computational efficiency, showing significant improvements over existing methods, such as LeGrad.
-
Jetson Inference: Hello AI World Project
Forked and customized NVIDIA’s jetson-inference project to run real-time object detection on NVIDIA Jetson Orin NX.
- Deployed the official Docker container and created a custom image with CUDA-enabled OpenCV, publishing it on Docker Hub.
- Integrated deep learning models with onboard camera input to demonstrate real-time AI applications on Linux for Tegra.
-
Attack Detection for Renewable Energy Systems Based on Cyclic Generative Networks
Developed a novel attack detection framework for renewable energy systems using cycleGANs to model normal operational patterns and identify anomalies indicative of cyber-attacks.
- Integrated LSTM modules into the cycleGAN architecture to capture temporal dependencies in time-series data.
- Utilized domain mapping and cycle-reconstruction loss to flag inputs exhibiting inconsistent round-trip transformation.
- Quantitatively evaluated anomaly by comparing reconstruction loss distributions between normal and abnormal data groups.
- Achieved over 92% AUC in distinguishing abnormal samples, validating the effectiveness of the cycleGAN-based approach.
-
Electricity Theft Detection Framework Using TSUNet and k-Shape Clustering
Proposed a scalable and generalizable electricity theft detection framework based on deep learning.
- Implemented adaptive k-shape clustering to select behaviorally diverse normal samples, balancing the dataset.
- Developed TSUNet, a time-series U-Net model, to capture multi-scale temporal features and generate localized time-step-level theft predictions.
- Validated the approach on both public (SGCC) and private U.S. utility datasets, achieving strong cross-domain generalization.
-
Signal Distortion Measuring Device
Design a signal distortion measuring device based on TI’s MCU. The device can measure the distortion of the signal and display the result on the screen.
- Designed hardware modules
- Derived signal processing algorithms
- Wrote code for MCU and FPGA
-
Cold Chain Temperature and Humidity Detection Tag based on Passive Backscatter Technology
Design transmitter based on backscatter technology and transmit RF signals without energy storage devices. The tag can be used in cold chain logistics to monitor the temperature and humidity of the goods.
- Search for circuit solutions for RF energy harvesting and design PCBs
- Wrote Verilog codes to compose message structure of bluetooth on FPGA
- Downloaded the program to the board, used a high-power RF signal source to provide the carrier, and used a packet sniffer to grab the Bluetooth packets
-
Digital Filter Design and Implementation on XILINX FPGA
Implemented an FIR filter on XILINX FPGA using Verilog and SystemVerilog, achieving real-time processing of audio signals.
- Generated parameters of the FIR filter using MATLAB’s Filter Designer (fdatool) to specify order, cutoff frequency, and window type for optimal audio signal processing.
- Optimized the filter implementation by restructuring the tap delay line and using fixed-point arithmetic, reducing LUT utilization by 35% and RAM usage by 40%
-
Circuit Checking System Based on ZYNQ
Designed and implemented a circuit checking system based on ZYNQ, enabling real-time monitoring and analysis of electrical circuits.
- Built PCBs for high-speed ADC (AD9226) and DAC (AD9767) chips to interface with ZYNQ.
- Developed FPGA logics in Verilog and package them into AXI-lite IP for ADC/DAC sampling and interfacing with ARM cores.
- Programmed the ARM core in C to acquire and transmit >100 MHz voltage signals from circuit probes.
- Identified fault types and locations based on frequency-domain attenuation; achieved 90% accuracy in fault localization.
References
- Professor Jeff (Jun) Zhang