About me

I am currently a Ph.D. student in the School of Computer Science at Georgia Tech. I work with Prof. Ling Liu and Prof. Calton Pu at the Distributed Data Intensive Systems Lab (DiSL). Before coming to Georgia Tech, I obtained my Bachelor’s degree with honors in Computer Science and Technology from University of Science and Technology of China (USTC).

My research interests include:

  • Systems for Machine Learning
  • Machine Learning for Systems
  • Big Data Systems & Analytics
  • Edge AI Systems

Publications

You can also find my articles on Google Scholar.

  • Boosting Ensemble Accuracy by Revisiting Ensemble Diversity Metrics
    Yanzhao Wu, Ling Liu, Zhongwei Xie, Ka-Ho Chow, and Wenqi Wei
    2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2021)
    [paper][code]

  • Gradient-Leakage Resilient Federated Learning
    Wenqi Wei, Ling Liu, Yanzhao Wu, Gong Su, and Arun Iyenger
    2021 IEEE International Conference on Distributed Computing Systems (ICDCS 2021) [paper][arXiv][code]

  • Learning TFIDF Enhanced Joint Embedding for Recipe-Image Cross-Modal Retrieval Service
    Zhongwei Xie, Ling Liu, Yanzhao Wu, Lin Li, Luo Zhong
    2021 IEEE Transactions on Services Computing
    [paper][arXiv][code]

  • Promoting High Diversity Ensemble Learning with EnsembleBench
    Yanzhao Wu, Ling Liu, Zhongwei Xie, Juhyun Bae, Ka-Ho Chow, Wenqi Wei
    2020 IEEE International Conference on Cognitive Machine Intelligence (IEEE CogMI 2020)
    [paper][arXiv][code][video]

  • Cross-Modal Joint Embedding with Diverse Semantics
    Zhongwei Xie, Ling Liu, Yanzhao Wu, Lin Li, Luo Zhong
    2020 IEEE International Conference on Cognitive Machine Intelligence (IEEE CogMI 2020)
    [paper][code]

  • Memory Abstraction and Optimization for Distributed Executors
    Semih Sahin, Ling Liu, Wenqi Cao, Qi Zhang, Juhyun Bae, Yanzhao Wu
    2020 International Conference on Collaborative Computing: Networking, Applications and Worksharing (IEEE CIC 2020)
    [paper]

  • Adversarial Deception in Deep Learning: Analysis and Mitigation
    Wenqi Wei, Ling Liu, Margaret Loper, Ka-Ho Chow, Mehmet Emre Gursoy, Stacey Truex, Yanzhao Wu
    2020 IEEE International Conference on Trust, Privacy and Security in Intelligent Systems and Applications (IEEE TPS 2020)
    [paper][code]

  • Adversarial Objectness Gradient Attacks in Real-time Object Detection Systems
    Ka-Ho Chow, Ling Liu, Margaret Loper, Juhyun Bae, Mehmet Emre Gursoy, Stacey Truex, Wenqi Wei, Yanzhao Wu
    2020 IEEE International Conference on Trust, Privacy and Security in Intelligent Systems and Applications (IEEE TPS 2020)
    [paper][code][video]

  • Efficient Orchestration of Host and Remote Shared Memory for Memory Intensive Workloads
    Juhyun Bae, Gong Su, Arun Iyengar, Yanzhao Wu, Ling Liu
    2020 International Symposium on Memory Systems (MEMSYS 2020)
    [paper][arXiv][code][video][slides]

  • Understanding Object Detection Through an Adversarial Lens
    Ka-Ho Chow, Ling Liu, Mehmet Emre Gursoy, Stacey Truex, Wenqi Wei, Yanzhao Wu
    2020 European Symposium on Research in Computer Security (ESORICS 2020)
    [paper][arXiv][code]

  • A Framework for Evaluating Client Privacy Leakages in Federated Learning
    Wenqi Wei, Ling Liu, Margaret Loper, Ka-Ho Chow, Mehmet Emre Gursoy, Stacey Truex, Yanzhao Wu
    2020 European Symposium on Research in Computer Security (ESORICS 2020)
    [paper][arXiv][code]

  • Cross-Layer Strategic Ensemble Defense Against Adversarial Examples
    Wenqi Wei, Ling Liu, Margaret Loper, Ka-Ho Chow, Emre Gursoy, Stacey Truex, Yanzhao Wu
    2020 International Conference on Computing, Networking and Communications (IEEE ICNC 2020)
    [paper][arXiv][code]

  • Demystifying Learning Rate Policies for High Accuracy Training of Deep Neural Networks
    Yanzhao Wu, Ling Liu, Juhyun Bae, Ka-Ho Chow, Arun Iyengar, Calton Pu, Wenqi Wei, Lei Yu, Qi Zhang
    2019 IEEE International Conference on Big Data (IEEE BigData 2019)
    [paper][arXiv][code]

  • Denoising and Verification Cross-Layer Ensemble Against Black-box Adversarial Attacks
    Ka-Ho Chow, Wenqi Wei, Yanzhao Wu, Ling Liu
    2019 IEEE International Conference on Big Data (IEEE BigData 2019)
    [paper][arXiv][code]

  • Deep Neural Network Ensembles against Deception: Ensemble Diversity, Accuracy and Robustness
    Ling Liu, Wenqi Wei, Ka-Ho Chow, Margaret Loper, Emre Gursoy, Stacey Truex, Yanzhao Wu
    16th IEEE International Conference on Mobile Adhoc and Sensor Systems (IEEE MASS 2019)
    [paper][arXiv][code]

  • A Comparative Measurement Study of Deep Learning as a Service Framework
    Yanzhao Wu, Ling Liu, Calton Pu, Wenqi Cao, Semih Sahin, Wenqi Wei, Qi Zhang
    2019 IEEE Transactions on Services Computing
    [paper][arXiv][code]

  • Memory Disaggregation: Research Problems and Opportunities
    Ling Liu, Wenqi Cao, Semih Sahin, Qi Zhang, Juhyun Bae, Yanzhao Wu
    39th IEEE International Conference on Distributed Computing Systems (ICDCS 2019)
    [paper][code]

  • Experimental Characterizations and Analysis of Deep Learning Frameworks
    Yanzhao Wu, Wenqi Cao, Semih Sahin, and Ling Liu
    2018 IEEE International Conference on Big Data (IEEE BigData 2018)
    [paper][code]

  • Benchmarking Deep Learning Frameworks: Design Considerations, Metrics and Beyond
    Ling Liu, Yanzhao Wu, Wenqi Wei, Wenqi Cao, Semih Sahin, and Qi Zhang
    38th IEEE International Conference on Distributed Computing Systems (ICDCS 2018)
    [paper][code]

  • CCAligner: a token based large-gap clone detector
    Pengcheng Wang, Jeffrey Svajlenko, Yanzhao Wu, Yun Xu and Chanchal K. Roy
    40th International Conference on Software Engineering (ICSE 2018)
    [paper][code]

Research Experience

  • Data-efficient Learning with DNN Ensembles
    Smart Decisions Team, Facebook, May 2021 – Aug 2021
    Mentor: Dr. Yin Huang
    Focus: Data Efficiency, Ensemble Learning
    Goal: Study the data efficiency of DNN ensemble models and design effective subsampling strategies to improve data efficiency for training ML models.

  • High performance Object Detection on Edge Devices
    Distributed Data Intensive Systems Lab, Georgia Tech, Aug 2020 – May 2021
    Supervisor: Prof. Ling Liu
    Focus: Deep Learning, Edge AI
    Goal: Design and implement an efficient framework for supporting various object detection models and achieving high performance on multiple edge devices.

  • Pipeline Parallelism for Deep Learning Recommendation Models
    AI System SW/HW Co-Design Team, Facebook Research, May 2020 – Aug 2020
    Mentor: Dheevatsa Mudigere
    Focus: Deep Learning, Pipeline Parallelism
    Goal: Apply pipeline parallelism into Facebook deep learning recommendation models to accelerate distributed recommendation model training.
    Achievement: PipeDLRM - an open-sourced software package built on top of DLRM and PyTorch.

  • High Accuracy and Robust Ensemble of Deep Neural Networks
    Distributed Data Intensive Systems Lab, Georgia Tech, Aug 2019 – May 2020
    Supervisor: Prof. Ling Liu
    Focus: Deep Learning, Edge AI
    Goal: Design and implement an ensemble framework to improve deep neural network accuracy and optimize inference robustness on GPUs and edge devices.
    Achievement: EnsembleBench - a holistic framework for promoting high diversity ensemble learning.

  • A Performance Study of Deep Learning with the High-performance Storage System
    Storage Systems Research Group, IBM Research, May 2019 - July 2019
    Mentors: Dr. Daniel Waddington, Dr. Luna Xu
    Focus: Storage Systems, Deep Learning Frameworks
    Achievement: Conducting a comprehensive performance analysis of the high-performance storage system with different storage backends, such as persistent memory and SSD, with popular deep learning workloads.

  • Semi-automatic Hyperparameter Tuning for Training Deep Neural Networks
    Distributed Data Intensive Systems Lab, Georgia Tech, Aug 2018 – May 2019
    Supervisor: Prof. Ling Liu
    Focus: Deep Learning Training, Performance Optimization
    Goal: Accelerate deep learning training and improve the training efficiency via semi-automatic hyper-parameter tuning.
    Achievement: LRBench - a semi-automatic learning rate tuning tool to enhance the deep neural network training efficiency and accuracy.

  • Accelerating Deep Learning with Direct-to-GPU Storage
    Storage Systems Research Group, IBM Research, May 2018 – Aug 2018
    Mentors: Amit Warke, Dr. Daniel Waddington
    Focus: Storage Systems, Deep Learning Frameworks
    Achievement: Integrating the Direct-to-GPU storage system into Caffe to obtain over 2$\times$ performance improvement by reducing the overhead of data transmission.

  • Experimental Analysis and Optimization of Deep Learning Frameworks
    Distributed Data Intensive Systems Lab, Georgia Tech, Aug 2017 – May 2018
    Supervisor: Prof. Ling Liu
    Focus: Deep Learning Systems, Performance Analysis
    Goal: Analyze the hyper-parameters and basic components of Deep Learning and optimize Deep Learning Frameworks by tuning data-related and hardware-related parameters. Achievement: GTDLBench - a performance benchmark of deep learning frameworks to measure and optimize mainstream deep learning frameworks.

Teaching

TA for:

  • Georgia Institute of Technology
    • CS6220 Big Data Systems and Analytics (Fall 2021)
    • CS6675/CS4675 Advanced Internet Computing Systems and Application Development (Spring 2018, Spring 2019, Spring 2020, Spring 2021)
    • CS6235/CS4220 Embedded Systems and Real-Time Systems (Fall 2018)
  • University of Science and Technology of China
    • CS1001A Computer Programming A (Fall 2015)

Guest lectures for:

  • Georgia Institute of Technology
    • CS6220 Big Data Systems and Analytics
      • Fall 2019: GTDLBench & LRBench
      • Fall 2020: Emulab
    • CS6675/CS4675 Advanced Internet Computing Systems and Application Development
      • Spring 2019: Emulab
      • Spring 2020: Emulab
      • Spring 2021: AWS + Google Colab

Reviewer

  • Conference: ICDE 2018, UCC 2018, BDCAT 2018, ICDCS 2019, WWW 2021
  • Journal: IEEE TKDE, ACM TOIT

Open-source Projects

  • DP-Ensemble: Leveraging FQ-diversity metrics to identify high diversity ensemble teams with high performance effectively.
  • PipeDLRM: Using pipeline parallelism for training deep learning recommendation models.
  • EnsembleBench: A set of tools for building high-quality ensembles for machine learning and deep learning models.
  • LRBench: A semi-automatic learning rate tuning tool to improve the deep neural network training efficiency and accuracy.
  • GTDLBench: A performance benchmark of deep learning frameworks to measure and optimize mainstream deep learning frameworks.
  • Comanche: Accelerating deep learning with Direct-to-GPU storage with a modified Caffe and DeepBench.
  • CCAligner: A token based code clone detector for detecting large-gap copy-and-paste source codes.
  • PRISM: Building the LTS and Game model checkers for PRISM, a widely applied model checker for system analysis.

Awards & Scholarship

  • College of Computing Student Travel Award, December 2020
  • Qualified for Men’s Singles in 2020 NCTTA South Regional Championships
  • Outstanding Graduate Award (USTC), April 2017
  • Fourth Place for 2016 ISC Student Cluster Competition, June 2016
  • Excellent Student Scholarship (Top 3%, USTC), 2015-2016
  • Leadership Scholarship, 2014-2015
  • The Third Prize for Electromagnetism Paper Competition, June 2014

Non-academic

  • Sports: When I’m not working in the lab, you may find me running or playing table tennis on the court at Georgia Tech.
    • Running: I finished the half-marathon in the 2015 Hefei International Marathon Competition.
    • Table Tennis: I am a member of GTTTA.
  • Literature & music: I love reading novels and listening to music.

Contact