Hello! I’m Hiro.
Hirofumi Shiba (Hiro) is a Ph.D. student supervised by Kengo Kamatani at the Institute of Statistical Mathematics (ISM), Tokyo, Japan.
My research is mainly in statistical computation and machine learning, and a particular focus is on Monte Carlo simulation algorithms. By making them more efficient, scalable, and automated, I aim to broaden the applications of statistics and machine learning.
Before pursuing my Ph.D., I studied Mathematics at the University of Tokyo.
Interests
- Monte Carlo methods
e.g., MCMC, SMC & PDMP - Bayesian Statistical Modelling
especially in Political Science & Biostatistics - Bayesian Neural Networks, Nonparametrics & Kernel Methods
Research
My research interests revolve around developing and analyzing random algorithms, by leveraging insights from their (scaling / continuous-time / homogenization) limit dynamics. From these perspectives, algorithms reveal their intrinsic properties, and my work aims to unify the understanding of various sampling algorithms under a common mathematical framework, i.e., flows on the space of probability measures \(\mathcal{P}(X)\).
Software
I maintain the Julia package PDMPFlux.jl, and also contribute to the R package YUIMA.
Qualifications
Ph.D. in Statistical Science, 2028 (expected)
Graduate University for Advanced Studies, SOKENDAI, Japan
B.Sc. in Mathematics, 2023
University of Tokyo, Japan
Industrial Activities
Cooperative Researcher, 2023.4 – present
RCAST, the University of Tokyo
Researching on AI safety and semiconductor policy.
Data Scientist, 2023.4 – present
PreMedica Inc., Tokyo, Japan
Consulting in the field of medical data analysis.