Zheng Zheng
PhD Student
Department of Computer Science
University of Texas at Arlington
"Understanding how structured information and optimization dynamics shape learning systems."
Contact Information
About Me
Welcome to my academic website. I am a Ph.D. student in Computer Science at the University of Texas at Arlington, advised by Prof. Junzhou Huang. My research lies at the intersection of biomedical machine learning, multi-modal representation learning, and optimization dynamics. I am particularly interested in how information is injected, aligned, and optimized across heterogeneous modalities in complex learning systems.
My recent work focuses on heterogeneous aligned fusion for survival prediction under missing modalities (MIDL), multi-modal computational pathology, and gradient conflict in multi-objective learning. I investigate how shared and task-specific parameter subspaces evolve during training, and how optimization dynamics affect representation stability in biomedical models.
More broadly, I am interested in understanding how large models—such as protein foundation models and large language models—encode and transfer structured information. I aim to explore principled methods for information injection and unsupervised representation learning in foundation-scale systems.
Previously, I worked on efficient deep learning architectures and recommendation systems, including a low-rank feature interaction method presented at ICML 2023. Mzhegey current focus is on developing a deeper theoretical and empirical understanding of learning dynamics in multi-modal and multi-objective settings.
Research Interests
- Multimodal Learning under Missing Modalities
- Computational Pathology & Biomedical AI
- Heterogeneous Aligned Fusion
- Multi-Objective Optimization & Gradient Conflict
- Information Injection in Foundation Models
- Representation Learning & Subspace Dynamics
- Large-Scale Protein & Language Models
Recent News
2026
Our paper "Heterogeneous Aligned Fusion for Survival Prediction with Missing Modalities" has been accepted to Medical Imaging with Deep Learning (MIDL 2026).
January 2025
Our paper "Multiple Abstraction Level Retrieve-Augment Generation" was released on arXiv.
October 2023
Our paper "FiBiNet++: Reducing Model Size by Low-Rank Feature Interaction Layer for CTR Prediction" was published at CIKM 2023.
December 2023
Completed M.S. in Computer Science at Brandeis University.