Tuorui Peng

Ph.D. candidate in Statistics

• NowPh.D. candidate in Statistics — Northwestern University
• FromShenzhen, Guangdong, China
• RouteSZSHS · 2013 & 2016 → Tsinghua · 2019 → Northwestern · 2023

Interests

Statistical theory under heavy tails and shape constraints — robust, minimax, and distribution-free methods in high dimensions.

• Robust statistics
• Minimax theory
• High-dimensional statistical inference
• Statistical learning theory

• Efficient Monotone Single-Index Estimation Beyond Elliptical Designs

  Oct 2025 — Present

  Northwestern University

  • — Designed a new algorithm that overcomes the distributional limitations of prior algorithms for monotone single-index estimation; implemented data-driven adaptive parameter selection.
  • — Built cross-language packages (R version + Python version over one shared native core, BLAS-accelerated kernels, unit-tested).
  • — Experiments over various complicated designs show a wide margin in recovering the index direction over established rank-correlation and dimension-reduction baselines.

• Robust Inference under Shape Constraints with Heavy-Tailed Noise

  Jan 2025 — Apr 2026

  Northwestern University

  • — Developed theory and minimax-optimal guarantees for robust mean estimation under heavy-tailed and adversarially contaminated noise with nonconvex (star-shaped) constraints; preprint under review at Bernoulli (arXiv:2604.05063).
  • — Follow-up work on the behavior near the breakdown point, exhibiting an interaction between the adversary and the geometry of the constraint set.
  • arXiv:2604.05063

• Distribution-Free Inference & NN Modeling for Gene Expression

  Dec 2022 — May 2023

  Tsinghua University

  • — Built an end-to-end Python pipeline for genotype simulation, Mixture Density Network training, and KL-divergence-based evaluation against the known true conditional law.
  • — Extended conformal prediction to CDF-valued targets via a Wasserstein-1 score, yielding distribution-free bands with valid 90% / 95% / 99% coverage; validated over data-size, dimensionality, causal-rate and error-model settings.

• PMC-Patients — Large-Scale Patient Summary Dataset & Retrieval Benchmark

  Dec 2021 — Jul 2022

  Tsinghua University

  • — Built components of the data-collection pipeline behind a peer-reviewed clinical dataset, converting PubMed Central case reports into structured patient records (section-level triggers, paragraph-level extractors, regex demographic parsing); a large-scale dataset: 167k patient summaries, 3.1M patient–article and 293k patient–patient annotations.
  • — Fine-tuned BERT-based dense retrieval encoders and benchmarked them against baselines.
  • — Evaluated on a held-out 5k-query split, over a corpus of 11.7M articles and 155k candidate patients; published in Scientific Data (Nature Portfolio).
  • Nature Scientific Data
  • Project website

1. 01

   Robust mean estimation under star-shaped constraints with heavy-tailed noise

   Peng, T., Prasadan, A., & Neykov, M. · Under review at Bernoulli. Preprint: arXiv:2604.05063 · 2026

2. 02

   A large-scale dataset of patient summaries for retrieval-based clinical decision support systems

   Zhao, Z., Jin, Q., Chen, F., Peng, T., & Yu, S. · Scientific Data, 10(1), 909 · 2023

• Honorable Mention — MCM/ICM

  Mathematical Contest in Modeling (COMAP), track A

  2022

• Scholarship for Academic Advance

  Tsinghua University

  2020

• Five-star Zijing Volunteer

  五星级紫荆志愿者 · Tsinghua University Volunteer Association

  2023

• Ma Yuehan Cup — Shooting

  第 64 届马约翰杯 · 10 m air rifle 4th & air gun 5th

  2021

• Note

  Statistics Note — A Brief Summary of Statistics Minor Courses

  Long-running course summary note for the statistics minor courses at Tsinghua, compiled in LaTeX with an indexed PDF.

  Read
• Note

  High Dimensional Statistics Note 2024-2025

  Reading notes on high-dimensional statistics, following Wainwright, Vershynin, Rigollet–Hütter and van Handel.

  Read