Zihao Ye

Hi, I’m Zihao Ye (叶子皓), which means “Luminous Leaf” in the cultural context of China. I am a Member of Technical Staff at Wood Wide AI, where I work on foundation models for structured and tabular data. I received my M.S. from Carnegie Mellon University, where I worked with Prof. Pradeep Ravikumar on tabular foundation models and structured representation learning.

Previously, I earned my B.S. through a joint program between QMUL and BUPT. During my undergraduate studies, I worked with Prof. Changjae Oh on computer vision and trustworthy AI.

My current research interests are representation learning theory and application, with a recent focus on structure-aware modeling for tabular and relational data. My previous experience can be found in my bio.

In addition to my research pursuits, I am also passionate about music. I play guitar 🎸 and sing.

news

Jan 26, 2026	Joined Wood Wide AI as a Member of Technical Staff, working on foundation models for structured data.
Jun 7, 2024	A paper on reference-guided image filtering (de-raining) was accepted to ICIP 2024! 🎉✨

selected publications

2026

Preprint

Tabular Numeric Stretch Transformation

Zihao Ye, Juyong Kim, Johnna Sundberg, and 2 more authors

Under review, 2026

Under review at NeurIPS 2026

ABSTRACT arXiv

We introduce the stretch transformation framework, which formulates numeric feature preprocessing for tabular data as an optimization problem over the smoothness of the target function. We study an unsupervised variant that uniformly redistributes feature density, and a supervised variant that optimizes target-aware transformations by minimizing the Dirichlet energy of the target function in the transformed space. We establish connections to piecewise linear encoding, empirical CDF transformation, and target encoding, and evaluate the framework across 38 datasets from the TALENT benchmark.

2024

ICIP

Improving Image de-raining Models using Reference-guided Transformers

Zihao Ye, Jaehoon Cho, and Changjae Oh

IEEE International Conference on Image Processing, 2024

ABSTRACT arXiv PDF Blog Code Poster

Image de-raining is a critical task in computer vision to improve visibility and enhance the robustness of outdoor vision systems. While recent advances in de-raining meth- ods have achieved remarkable performance, the challenge remains to produce high-quality and visually pleasing de- rained results. In this paper, we present a reference-guided de-raining filter, a transformer network that enhances de- raining results using a reference clean image as guidance. We leverage the capabilities of the proposed module to further refine the images de-rained by existing methods. We vali- date our method on three datasets and show that our module can improve the performance of existing prior-based, CNN- based, and transformer-based approaches.