Xinrong Zhou
Hi there👋! I'm currently an undergraduate at The Chinese University of Hong Kong. I have a deep interest and firm grounding in Statistics and Computer Science, and I’m self‑motivated to keep pushing my limits. Glad you're here.
Education
- Bachelor of Science, major in Statistics; double major in Computer Science (First-Class honours), CUHK
Honors and awards
- Best Original Research Award, HKSSH Free Paper Session, 37th HKSSH Annual Congress (Combined APOA Hand & Upper Limb Society 2nd Meeting), Hong Kong, 2025
- HKSAR Government Scholarship (HK$80,000)
- Asia-Pacific Economic Cooperation Scholarship (APEC Honorary Award)
- Cheung Chuk Shan Scholarship, CUHK (HK$42,100)
- Tsang Shin Tim Scholarship, CUHK (HK$42,100)
- Fan Fang Qi Ying Memorial Scholarship, CUHK (HK$5,000)
- Tung Wah Group of Hospitals Prize (HK$3,000)
- Dean's List, CUHK (2024-2025, 2023-2024, 2022–2023)
- United College Head's List (Rank #1 of College in major Statistics), CUHK (2024-2025, 2023-2024, 2022-2023)
Research
Upper limb necrotizing fasciitis: 20‑year global study
This research provided an epidemiological study. Statistical association between the response variables and the predictors were studied, we also produced a practical scoring system to predict mortality with a simple calculation step, good predictive power, and intuitive interpretation.
Directional dependence statistic based on Chatterjee’s correlation
This is an ongoing theoretical research project aiming to establish the foundational properties of a novel directional statistic, Δq = q(X,Y) − q(Y,X), derived from Chatterjee's rank correlation. We aim to rigorously analyze its consistency, key invariance properties, and asymptotic behavior (LLN/CLT) under conditions of mild dependence. Stay tuned to see what we discover!
Projects
Robust animal re-identification (Kaggle/AnimalCLEF25)
This project tackled open‑set wildlife individual re‑identification. I led data‑split design, built an data augmentation pipeline, handled class imbalance, and fine‑tuned pre-trained MegaDescriptor with cross‑entropy and triplet losses. A broad grid search plus species‑specific thresholding delivered 110%–140% gains in GEO_MEAN.
Statistics capstone: generating realistic faces
This project built realistic face generators with VAE, GAN, and a hybrid approach. I implemented the pipelines and built a VAE–GAN hybrid optimized with a composite objective using coordinated encoder/decoder/ discriminator updates and carefully tuned hyperparameters; evaluated with Inception Score and Laplacian variance, then investigated latent‑space structure for facial‑feature interpolation.
AI image generator platform
This project delivered an end‑to‑end image generation web app. I co‑designed the architecture, built Flask REST APIs and a React UI, added authentication and an SQLite data layer, and wrote integration tests. The system supports SDXL generation and Real‑ESRGAN upscaling.
Skills
- XGBoost
- Gradient Boosting
- Random Forest
- VAEs
- GANs
- LSTM
- SVM
- Clustering
- Web scraping
- OOP
- LaTeX
- SPSS
- SAS
