Nov 1st, 2022

Xun Jiao

Assistant Professor

Villanova University

Email:

xun.jiao@villanova.edu

Personal webpage

https://vu-detail.github.io/people/jiao

Research interests

Robust Computing, Efficient Computing, AI/Machine Learning, Brain-inspired Computing, Fuzz Testing

Short bio

Xun Jiao is an assistant professor in ECE department of Villanova University. He leads the Dependable, Efficient, and Intelligent Computing Lab (DETAIL). Before that, he obtained his Ph.D. degree from UC San Diego in 2018. He earned a dual first-class Bachelor degree from Joint Program of Queen Mary University of London and Beijing University of Posts and Telecommunications in 2013. His research interests are on robust and efficient computing for intelligent applications such as AI and machine learning. He published 50+ papers in international conferences and journals. He received 6 paper awards/nominations in international conferences such as DATE, EMSOFT, DSD, and SELSE. He is an associate editor of IEEE Trans on CAD, and a TPC member of DAC, ICCAD, ASP-DAC, GLSVLSI, LCTES. His research is sponsored by NSF, NIH, L3Harris, and Nvidia. He has delivered an invited presentation at U.S. Congressional House. He is a recipient of 2022 IEEE Young Engineer of the Year Award (Philadelphia Section).

Research highlights

Robust computing
• With continuous scaling of CMOS technology, circuits are even more susceptible to timing errors caused by microelectronic variations such as voltage and temperature variations, making them a notable threat to circuit/system reliability. Dr. Jiao has adopted a cross-layer approach (circuit-architecture-application) to combat errors/faults originated in hardware. Specifically, Dr. Jiao has pioneered in developing machine learning-based models to model/predict the errors in hardware and take proactive actions such as instruction-based frequency scaling to prevent errors. By exploiting the application-level error resilience of different applications (e.g., AI/machine learning, multimedia), Dr. Jiao has also developed various approximate computing techniques for more efficient execution.

Energy-efficient computing
• Energy efficiency has become a top priority for both high-performance computing systems and resource-constrained embedded systems. Dr. Jiao proposed solutions to this challenge at multiple abstraction levels. He proposed intelligent dynamic voltage and frequency scaling (DVFS) for circuits and systems, as well as designing novel efficient architecture such as in-memory computing and bloom filter to execute emerging workloads such as deep neural networks.

AI/brain-inspired computing
• Hyperdimensional computing (HDC) was introduced as an alternative computational model mimicking the “human brain” at the functionality level. Compared with DNNs, the advantages of HDC include smaller model size, less computation cost, and one/few-shot learning, making it a promising alternative computing paradigm. Dr. Jiao’s work has been pioneering the robustness of HDC against adversarial attacks and hardware errors, which has earned him a best paper nomination at DATE 2022. He also applied HDC to various application domains such as natural language processing, drug discovery, and anomaly detection, which demonstrated promising performance compared to traditional learning methods.

Fuzzing-based secure system
• Cyber-security in the digital age is a first-class concern. The ever-increasing use of digital devices, unfortunately, is facing significant challenges, due to the serious effects of security vulnerabilities. Dr. Jiao has developed a series of vulnerability detection techniques based on fuzzing, and has applied to software, firmware, and hardware. Over 100 previously unknown vulnerabilities are discovered and are reported to the US National Vulnerability Database with unique CVE assignments. He received two best paper nominations from EMSOFT 2019 and 2020.