Mathematical statistician advancing asymptotic inference for Riemannian optimization and local differential privacy.
2
papers here
2
reviewed by Evolve
2
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.
Research profile
Output—
No publication-volume data is available in this snapshot, so output cannot be scored. This axis would measure the count and cadence of peer-reviewed papers, venues, and authorship roles over time.
Impact—
Citation and influence metrics are not yet available for this researcher, so impact is unscored. This axis would measure citations, h-index-style signals, and uptake of the two inference methods by other works.
Focus0/100
The radar assigns focus a score of 0, derived from a field distribution split exactly evenly (0.5/0.5) between two labels: mathematical statistics with differential privacy and Riemannian optimization, and nonparametric inference on Riemannian manifolds. In substance the two catalogued papers are tightly related — both build asymptotic or simultaneous inference for manifold-valued estimation — so the low score reflects label breadth rather than genuinely scattered interests.
Collaboration36/100
The collaboration dimension scores 36, based on 5 distinct coauthors observed across the catalogued work. The recurring partners are Yangdi Jiang and Qirui Hu, appearing on both papers, which points to a small, stable working group rather than a broad network. Both papers list the researcher as first author, suggesting a coauthorship pattern centered on a compact team producing theory-heavy manuscripts.
Activity—
Temporal activity data is insufficient here, so activity cannot be scored. This axis would measure publication cadence and recent output; the only dated evidence is two arXiv papers posted in 2026, too thin to establish a trend.
Reproducibility55/100
Reproducibility scores 55, computed as the mean readiness over the 2 catalogued papers, each rated medium. Both works are theory-first with formal asymptotic guarantees, and Evolve's reviews note practical caveats: one relies on per-design safeguard tuning for coverage, and the other lacks baseline comparisons. Reproducing the numerical results would therefore require careful reimplementation of the stated tuning and calibration procedures.
Research summary
Xiaotian Chang is affiliated with Nanyang Technological University and works in mathematical statistics; this profile is name-resolved from public bibliographic data. The Evolve catalog contains two 2026 arXiv papers, both with Chang as first author, sitting in two evenly weighted (0.5/0.5) field areas: mathematical statistics with differential privacy and Riemannian optimization, and nonparametric inference on Riemannian manifolds. The first paper, 'Locally Private Inference for Riemannian Stochastic Optimization,' develops central limit theorems and transcript-based Wald inference for manifold-valued minimizers under local differential privacy, using Polyak-Ruppert averaging; Evolve judges the work rigorous while noting that coverage relies on per-design safeguard tuning. The second, 'Riemannian Simultaneous Inference for Tangent Vector Field Regression,' constructs Gumbel-calibrated simultaneous confidence tubes for kernel regression of tangent vector fields via parallel transport; the theory is rated strong, with small-sample coverage flagged as fragile and baseline comparisons absent. Across the catalogued work, 5 distinct coauthors appear, with Yangdi Jiang and Qirui Hu recurring on both papers, and the radar collaboration score is 36. Reproducibility scores 55, the mean readiness across the two papers, each rated medium. Output, impact, and activity metrics are not yet available in this snapshot, so those dimensions remain unscored rather than estimated. Overall, this is a focused, theory-driven early profile centered on rigorous inference for manifold-valued and privacy-constrained estimation.
Collaboration pattern
The catalogued papers show a compact collaboration pattern: 5 distinct coauthors in total, with Yangdi Jiang and Qirui Hu recurring on both 2026 papers. Chang is first author on each, and neither lists a corresponding-author role, consistent with a small theory-oriented team rather than a distributed project. The radar collaboration score is 36, in the lower range, reflecting limited observed network breadth in this snapshot.
Research keywords
Highlights
Local DP inference for manifold-valued stochastic optimization
Gumbel-calibrated simultaneous bands for tangent field regression
Timeline
- 2026Posts first-author arXiv paper 'Locally Private Inference for Riemannian Stochastic Optimization' on central limit theorems and Wald inference under local differential privacy.
- 2026Posts first-author arXiv paper 'Riemannian Simultaneous Inference for Tangent Vector Field Regression' on Gumbel-calibrated simultaneous confidence tubes.
- 2026Both 2026 papers receive medium reproducibility readiness in Evolve reviews, averaging to a reproducibility score of 55.
Papers in the Evolve catalog
Frequent co-authors
About this profile
This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.
- Bibliometric signals come from a limited snapshot; the output, impact, and activity dimensions lack underlying data and are not scored.
- No ORCID or country information was available, so the profile is name-resolved from public bibliographic data and same-name collisions cannot be fully ruled out.
- Coverage is limited to two 2026 arXiv papers in the Evolve catalog; peer-reviewed publications, other preprints, and non-English work may be missing.
- Radar scores (collaboration 36, focus 0, reproducibility 55) are Evolve-computed metrics and should be read alongside the small two-paper sample.