Fengxiang He

Algorithm Scientist
JD Explore Academy
JD.com Inc
hefengxiang@jd.com
fengxiang.f.he@gmail.com

Bio

I recieved BSc in statistics from University of Science and Technology of China and MPhil and PhD in computer science from University of Sydney supervised by Professor Dacheng Tao. I am currently algorithm scientist and DMT management trainee at JD Explore Academy, JD.com Inc. leading the Trustworthy AI Group. My research interest is in the theory and practice of trustworthy AI, including deep learning theory, privacy-preserving ML, decentralized learning, and algorithmic game theory.

The Trustworthy AI Group is recruiting research interns, algorithm scientists, and algorithm engineers.

Google Scholar | DBLP | Semantic Scholar | LinkedIn

Featured Publications

See full publication list here. Any comment is welcome.

Research papers

  1. Mu Yuan, Lan Zhang, Fengxiang He, Xueting Tong, Xiang-Yang Li. "InFi: End-to-End Learning to Filter Input for Resource-Efficiency in Mobile-Centric Inference." arXiv, 2022. (Full version of "InFi: End-to-end Learnable Input Filter for Redundancy-measurable Inference Workload." MOBICOM, 2022.) [paper] [code] [bib] NEW

  2. Yikai Wang, Wenbing Huang, Fuchun Sun, Fengxiang He, and Dacheng Tao. “Channel Exchanging Networks for Multimodal and Multitask Dense Image Prediction.” IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2022. [paper] [bib] NEW

  3. Tian Qin*, Fengxiang He*, Dingfeng Shi, Wenbing Huang, and Dacheng Tao. “Benefits of Permutation-Equivariance in Auction Mechanisms.” Advances in Neural Information Processing Systems (NeurIPS), 2022. [paper] [bib] NEW

  4. Kaiwen Yang, Yanchao Sun, Jiahao Su, Fengxiang He, Xinmei Tian, Furong Huang, Tianyi Zhou, and Dacheng Tao. “Adversarial Auto-Augment with Label Preservation: A Representation Learning Principle Guided Approach.” Advances in Neural Information Processing Systems (NeurIPS), 2022. [paper] [bib] NEW

  5. Tianying Ji, Yu Luo, Fuchun Sun, Mingxuan Jing, Fengxiang He, and Wenbing Huang. “When to Update Your Model: Constrained Model-based Reinforcement Learning.” Advances in Neural Information Processing Systems (NeurIPS), 2022. [paper] [bib] NEW

  6. Tongtian Zhu, Fengxiang He, Lan Zhang, Zhengyang Niu, Mingli Song, Dacheng Tao. "Topology-aware Generalization of Decentralized SGD." International Conference on Machine Learning (ICML), 2022. [paper] [code] [bib]

  7. Shaopeng Fu*, Fengxiang He*, Dacheng Tao. "Knowledge Removal in Sampling-based Bayesian Inference." International Conference on Learning Representation (ICLR), 2022. (Part of Shaopeng Fu*, Fengxiang He*, Dacheng Tao. "Bayesian Inference Forgetting.") [paper] [code] [bib]

  8. Fengxiang He*, Bohan Wang*, and Dacheng Tao. "Tighter generalization bounds for iterative differentially private learning algorithms." Conference on Uncertainty in Artificial Intelligence (UAI), 2021. [paper] [bib]

  9. Fengxiang He*, Bohan Wang*, and Dacheng Tao. "Piecewise linear activations substantially shape the loss surfaces of neural networks." International Conference on Learning Representation (ICLR), 2020. [paper] [website] [poster] [bib]

  10. Fengxiang He, Tongliang Liu, and Dacheng Tao. "Control batch size and learning rate to generalize well: Theoretical and empirical evidence." Advances on Neural Information Processing (NeurIPS), 2019. [paper] [poster] [bib]

White papers

  1. Blue Report on the Industrial Ecosystem of Trustworthy Artificial Intelligence
    China Academy of Information and Communications Technology and JD Explore Academy Leading author

  2. White Paper on AIoT in Healthcare: A Corporate Blueprint in China and India
    BRICS Institute of Future Networks, China Academy of Information and Communications Technology, Indian Institute of Technology, International Institute of Information Technology, and JD Explore Academy
    Leading author of Chapter 5 "Vision and future possibilities for AIoT in healthcare"
    [English version]

  3. White Paper on Trustworthy Artificial Intelligence
    JD Explore Academy and China Academy of Information and Communications Technology
    Co-leading author
    [English version] [Chinese version]

* Co-first authors.

Trustworthy AI Group

The Trustworthy AI group at JD Explore Academy endeavours to pursue excellence in research and development of trustworthy AI including

The Trustworthy AI Group is recruiting research interns and algorithm engineers.

Trustworthy AI Consulting Services

The Trustworthy AI Group works closedly with partners and clients to design ethics-aware AI systems based on the knowledge and advances of our cutting-edge research and development programs. The services are delivered on our JD Trustworthy AI Assessment Platform. Our partners and clients include major players from the retail and autonomous driving sectors.

Trustworthy AI Assessment Platform

The JD Trustworthy AI Assessment Platform is designed and developed by the Trustworthy AI Group at JD Explore Academy. It assesses AI models from over 20 aspects on the privacy-preserving, robustness, and fairness properties. It has been deployed within the JD.com Inc. to secure the trustworthiness of AI products and services. The Platform will also be deployed in the assessment services, corperately delivered with the Chinese Academy of ICT.

OmniForce: Privacy Computing

OmniForce is the autoML platform developed by JD Explore Academy, featured by the high compatibility with metaverse and large-scale foundation models. The Trustworthy AI Group works closely with the OmniForce team and external partners and clients to develop the Privacy Computing modules in OmniForce. Our clients and partners include major players from the telecommunications and education sectors.

Talks

Last update: Fri 30 Sep 2022