Ph.D. Statistics (Rutgers University).
News:
「AI時代の統計学」という研究集会を主催しています。ウェブサイトはこちら
I have released my new probability paper "Sharp Bounds for Near-Maximizers of Gaussian Processes," on arXiv.
My first-authored paper on uncertainty quantification in LLMs has been accepted at Neurips2026! "CITE: Anytime Valid Statistical Inference in LLM Self-Consistency" (joint work with N. Iwase, Y. Ichihara, J. Komiyama and M. Imaizumi).
I will present my new working paper "Poisson Geometry of Selective Inference in Proportional Dimensions" at EcoSta 2026. WP in preparation.
I have released my new paper "Finite-Sample Inference for Sparsely Permuted Linear Regression" (joint work with Masaaki Imaizumi), revision requested by J. Amer. Stat. Assoc.
Affiliations:
Project Assistant Professor at Komaba Institute for Science, University of Tokyo, Department of Basic Science, Graduate School of Arts and Science (Imaizumi-lab)
Visiting Scientist at RIKEN Center for Advanced Intelligence Project, High-Dimensional Structure Theory Team
Curriculum Vitae:
Before joining the University of Tokyo/RIKEN, I completed my Ph.D. in Statistics at Rutgers University (USA) and I am very fortunate to be advised by Professor Min-ge Xie.
E-mail: hirofumi-ota (at) g.ecc.u-tokyo.ac.jp
(previous email address "ho105 (at) stat.rutgers.edu" and "ho105 (at) rutgers.edu" are no longer in use).
Research interest: mathematical statistics, econometrics, applied probability and machine learning.
My research interest lies in mathematical foundations of statistical inference, data science and artificial intelligence.
Especially, my research is driven by problems in modern data science that resist existing uncertainty quantification frameworks. I develop new theory and methods that broaden the class of models and data structures for which reliable parameter inference can be achieved.
(i) modern uncertainty quantification
selective inference
irregular inference problems (discrete, non-numerical parameters)
repro samples method, universal inference, conformal inference, likelihood-free inference, etc.
statistical guarantees for reliable reasoning in LLMs/LLM alignment.
(ii) E-values/E-processes
theory, method and applications to modern ML
sequential inference
time series monitoring
(iii) econometrics
quantile models
non-standard asymptotics
(iv) mathematical statistics
high-dimensional/nonparametric statistics
information aggregation
shape-constrained estimation and inference
(v) applied probability
randomization/permutation
distributional approximation
Martingales
Honors and Awards:
Best presentation award at the Conference of Japan Statistical Society, 2025
Truman-Koehler Scholarship/Excellence Award (Rutgers University, USA)
Grant-in-Aid for JSPS Fellows (DC1, Japan)
Reviewing services:
Electronic Journal of Statistics, Journal of Computational and Graphical Statistics, Journal of Time Series Analysis, Annals of Institute of Statistical Mathematics, Econometric Reviews, Scandinavian Journal of Statistics, Journal of Business & Economic Statistics.
Other professional experiences:
Office of Statistical Consulting (Department of Statistics, Rutgers University) -- consulting services on statistical theory and methodology for companies, researchers, and graduate students.