Xiangyu Zhang
Xiangyu Zhang

Xiangyu Zhang

Ph.D. Candidate in Statistics

University of Minnesota

Email | CV | Research | Software

Bio

I am a sixth-year Ph.D. candidate in the School of Statistics at the University of Minnesota, advised by Prof. Sara Algeri and co-advised by Prof. Charlie J. Geyer. I am mostly interested in distribution-free statistical inference, Large Language Models and their applications, and applied ML and statistical inference for scientific domains, especially astronomy.

I combine rigorous statistical methodology with practical machine learning implementation, with a focus on reliable inference for scientific data, computationally efficient goodness-of-fit testing, and reproducible research software. Before beginning my Ph.D., I completed undergraduate studies in Mathematics and Statistics, with a minor in Computer Science, at the University of Minnesota.

Selected Publications

DF

A New Class of Asymptotically Distribution-Free Smooth Tests

Xiangyu Zhang and Sara Algeri.

Under review. 2026.

Develops asymptotically distribution-free smooth tests that remain valid with estimated parameters, model selection, and moderate sample sizes.

[arXiv] [Details]

APS

Validating Angular Power Spectral Models for the Stochastic Gravitational-Wave Background Without Distributional Assumptions

Xiangyu Zhang, Erik Floden, Hongru Zhao, Sara Algeri, Galin Jones, Vuk Mandic, and Jesse Miller.

Physical Review D 113, 023051. 2026.

Extends distribution-free inference to estimation and testing of SGWB angular power spectral models when the estimator likelihood is unavailable or poorly approximated by a Gaussian form.

[arXiv] [DOI] [Details]

GOF

Testing Models for Angular Power Spectra: A Distribution-Free Approach

Sara Algeri, Xiangyu Zhang, Erik Floden, Hongru Zhao, Galin Jones, Vuk Mandic, and Jesse Miller.

Physical Review D 113, L021306. 2026.

Introduces a goodness-of-fit strategy for angular power spectra with unknown parameters, avoiding distributional assumptions and repeated case-by-case simulations.

[arXiv] [DOI] [Details]

RT

A Novel Approach to Detect Line Emission Under High Background in High-Resolution X-Ray Spectra

Xiangyu Zhang, Sara Algeri, Vinay Kashyap, and Margarita Karovska.

Monthly Notices of the Royal Astronomical Society 521(1), 969-983. 2023.

Develops a statistical approach for detecting emission features or setting upper limits in high-background spectra, with application to Chandra observations of RT Cru.

[arXiv] [DOI] [Details]

LP

Exhaustive Goodness-of-Fit via Smoothed Inference and Graphics

Sara Algeri and Xiangyu Zhang.

Journal of Computational and Graphical Statistics 31(2), 378-389. 2022.

Builds smoothed-inference tools and graphics for exhaustive goodness-of-fit assessment, forming part of the statistical foundation behind LPsmooth.

[DOI] [Details]

Experience

The Travelers Companies, Inc.

Data Scientist Summer Intern, Data Science Leadership Development Program. St. Paul, MN. Jun. 2025 - Aug. 2025.

Developed LightGBM predictive models with Tweedie loss for 96-month general liability losses and expenses from early-stage claim and claimant attributes. Built Python data pipelines for claims data cleaning, feature transformation, and bucketing.

Validated models with cross-validated RMSE, SHAP importance, and lift charts against actuarial baselines, then presented business insights and implementation recommendations to 200+ colleagues.

University of Minnesota

Research Scientist, Distribution-Free Testing Framework. Minneapolis, MN. Jun. 2021 - May 2026.

Developed and evaluated distribution-free testing methods for complex and noisy data settings, with applications to gravitational-wave model validation and binary-star signal detection.

Produced reproducible code and research outputs across five publications in statistical inference and astrostatistics.

University of Minnesota

Teaching Assistant. Sep. 2020 - Jan. 2026.

Served as a teaching assistant for Statistical Machine Learning for more than three years, mentoring student cohorts on end-to-end ML projects covering preprocessing, feature engineering, model selection, and performance evaluation.

Also supported courses in statistical theory and advanced applied statistics, with teaching evaluations above 5.5/6.

Software

R

LPsmooth

R package for goodness-of-fit testing with smooth tests and comparison density plots.

[CRAN]

Py

DisfreeTestAPS

Python repository for distribution-free goodness-of-fit tests with tutorials and examples.

[Code]

Py

LPBkg

Python package for statistical modeling under background mismodeling scenarios.

[PyPI]

Contact