Zhengquan Luo

Department of Automation, University of Science and Technology of China

About Me

I am a Ph.D. candidate at the University of Science and Technology of China, specializing in federated learning, privacy-preserving computation, and distributed biometric systems.

Education

• Ph.D. in Engineering, University of Science and Technology of China (Aug 2018 - Present)
• M.A. in Automation, University of Science and Technology of China (Aug 2018 - Jun 2020)
• B.Eng. in Automation, University of Science and Technology of China (Aug 2014 - Jun 2018)

Work Experience

• Research Assistant, Ant Group (Alibaba Corporation) (Jul 2022 - Present)

Research Interests and Results

• Generalization theory and optimization algorithms for heterogeneous federated learning
• Privacy-preserving computation and data security for distributed biometric features
• Joint optimization theory and personalization for distributed biometric LLMs
• 8 papers published at conferences/journals
• 5 authorized patents and 3 published PCTs as the first inventor
• Key contributor to significant projects

Projects and Funds

• Federated Learning for Biometrics Recognition (CAAI-Huawei Mindspore Open Fund, Jan 2022 - Oct 2022)
• National Natural Science Foundation of China (No.62176246, 61836008, 62006225, 61906199, 62071468)
• Strategic Priority Research Program of Chinese Academy of Sciences (CAS) (Grant No. XDA27040700)

Awards

• PhD period:
  • Postgraduate National Scholarship (first class2, second class2)
• BS period:
  • The first prize of the provincial electronic design competition in Anhui Province, China
  • National Scholarship (Gold3, Silver1)
  • Excellence Scholarship

Publications

Journal

• Federated Local Compact Representation Communication: Framework and Application
• Combining 2D texture and 3D geometry features for Reliable iris presentation attack detection using light field focal stack
• Balanced Representation Learning for Long-tailed Skeleton-based Action Recognition
• SplitFedGaze: Distributed, Privacy-preserving, and Personalized Gaze Estimation (TMM under review)

Conference Full Papers

• Disentangled Federated Learning for Tackling Attributes Skew via Invariant Aggregation and Diversity Transferring
• FedIris: Towards More Accurate and Privacy-preserving Iris Recognition via Federated Template Communication
• Iris Normalization Beyond Appr-Circular Parameter Estimation
• Consistency-constrained Interactive Text-to-image Generation (preparing submit to NeurIPS)
• Metaverse Gaze Privacy Protection Based on Split Federated Learning (under review)
• PDVN: A Patch-based Dual-view Network for Face Liveness Detection using Light Field Focal Stack
• A Large-scale Database for Less Cooperative Iris Recognition

Patents

• Iris image feature extraction method, system and device based on federated learning
• A method of forming iris normalized images
• A federal face image feature learning method
• Disentanglement personalized federated learning method for consensus representation extraction and diversity propagation
• An updated method for node models that resists the spread of discrimination in federated learning
• IRIS IMAGE FEATURE EXTRACTION METHOD AND SYSTEM BASED ON FEDERATED LEARNING (International Application PCT/CN2021/092794)
• DISENTANGLED PERSONALIZED FEDERATED LEARNING METHOD FOR CONSENSUS REPRESENTATION EXTRACTION AND DIVERSITY PROPAGATION (International Application PCT/CN2022/135821)
• NODE MODEL UPDATING METHOD FOR RESISTING BIAS TRANSFER IN FEDERATED LEARNING (International Application PCT/CN2022/135819)

Professional Activity

• Reviewing Conferences: ICML, CVPR, NeurIPS, AAAI, IJCB, CCBR

Contact Information

• Email: [email protected]
• Address: No.95 ZhongGuanCun East Street, HaiDian District, Beijing, P.R. China, 100190

Last updated: March 14, 2024