AccelerateZ's Site

Appearance

Duoxiang Zhao ​

Student

Major in Computer Science at Sichuan University - Pittsburgh Institute (SCUPI)

Education ​

  • Sichuan University, Chengdu, China (September 2023 - June 2027, estimated)
    • Major in Computer Science at the Pittsburgh Institute (SCUPI)

Achievements ​

  • 2025-09 - Provincial Second Prize in_National Undergraduate Mathematical Contest in Modeling
  • 2025-09 - The Second Prize Scholarship, Sichuan University (2024-2025 Academic Year)
  • 2025-09 - Outstanding Student Leader, Sichuan University (2024-2025 Academic Year)
  • 2024-09 - The Third Prize Scholarship, Sichuan University (2023-2024 Academic Year)
  • 2022-09 - Second Prize in National High School Mathematics League (Preliminary)

Skills ​

Natural Languages ​

  • English (CET 6)
  • Chinese (Native)

Programming Languages ​

  • C and C++ - Algorithm and systems programming
  • Java - Object-oriented development
  • Python - Data science and deep learning
  • JavaScript - Frontend development

Other Skills ​

  • CUDA
  • Ubuntu
  • Docker
  • Markdown
  • LaTeX

Publications ​

  • Xing Wei, Duoxiang Zhao, et al. "GPG-HT: Generalized Policy Gradient with History-Aware Decision Transformer for Probabilistic Path Planning," under review, 2025.
  • Duoxiang Zhao, et al. "Conceptor-Augmented Regularization for RNN and LSTM Networks," in Proc. 2025 IEEE 2nd International Conference on Deep Learning and Computer Vision (DLCV), 2025. doi: 10.1109/DLCV65218.2025.11088507
    Abstract

    Recurrent Neural Network (RNN) and its variants, such as Long Short-Term Memory (LSTM), have achieved remarkable success in sequential data processing tasks. The representations learned by their hidden layers often require further regularization to enhance model performance and robustness. In this paper, we propose a novel Conceptor-based regularization mechanism for improving the hidden state representations of RNN and LSTM networks. Unlike previous works that primarily treat Conceptors as classifiers, we introduce a dynamic Conceptor layer integrated within the hidden states of RNNs and LSTM networks, termed Conceptor-Augmented RNN (CA-RNN) and Conceptor-Augmented LSTM (CA-LSTM). We evaluate the proposed method on the task of classifying speakers using Japanese vowel recordings, comparing standard RNN and LSTM models with those augmented by our Conceptor-based regularization. Our results demonstrate that incorporating the proposed dynamic Conceptor mechanism significantly improves classification performance and noise resistance, highlighting its effectiveness as a hidden-state regularizer for sequential models.

Research Experiences ​

Research Student at SCUPI CS Laboratory ​

Details
  • Sichuan University - Pittsburgh Institute (May 2025 - present)
  • Advisors: Prof. Guangwu Qian, Prof. Yuqi Ouyang
  • Focus on deep learning applications, feedback networks, and video anomaly detection
  • Maintained lab servers and deployment pipelines; gained familiarity with distributed infrastructure

Honor Undergraduate Research ​

Details
  • Sichuan University - Pittsburgh Institute (April 2024 - April 2025)
  • Learned the full research workflow, from literature review to experimentation
  • Practiced software tooling for academic research and inspected state-of-the-art methods

Modeling Optimization at West China Hospital ​

Details
  • Sichuan University - West China Hospital (Sept. 2024 - Oct. 2024)
  • Applied convolutional neural networks for visual classification of pressure ulcer stages
  • Mitigated overfitting via data augmentation and algorithmic refinements
  • Improved model implementations for robustness

Teaching Experiences ​

Teaching Assistant for INFSCI 0510: Data Analysis ​

Details
  • Sichuan University - Pittsburgh Institute (March 2026 - June 2026)
  • Supported an introductory machine learning course: logistic regression, SVM, decision trees, KNN, Naive Bayes, PCA, K-Means, DBSCAN, GMM, EM, ensemble methods, and evaluation
  • Held lab hours and graded assignments and exams