☕️About Me
I’m Hiroki Sakamoto, a Ph.D. candidate in Information Science and Technology at the University of Tokyo. I am a member of Mathematical Informatics Laboratory No. 5, advised by Kazuhiro Sato.
My research focuses on model reduction and compression for high‑dimensional dynamical systems and deep learning models. Broadly, I explore data‑driven methods for modeling, controlling, and interpreting complex physical systems, aiming to create compact yet accurate representations that bridge control theory and modern machine‑learning techniques.
Interests
- Model Order Reduction
- Model Compression
- Randomized Numerical Linear Algebra
Education
Ph.D. Candidate in Mathematical Informatics
The University of Tokyo
Exchange Student
Sorbonne University
M.S. in Mathematical Informatics
The University of Tokyo
B.S. in Mathematics
Waseda University
Recent Publications
Publications
(2025).
A Deep State-Space Model Compression Method using Upper Bound on Output Error.
preprint on arXiv.
(2025).
Compression Method for Deep Diagonal State Space Model Based on H2 Optimal Reduction.
IEEE Control Systems Letters, Vol. 9, pp. 2043–2048.
(2025).
Data-driven h2 model reduction for linear discrete-time systems.
conditionally accepted for publication in IEEE Transactions on Automatic Control.
(2024).
Stable Linear System Identification with Prior Knowledge by Riemannian Sequential Quadratic Optimization.
IEEE Transactions on Automatic Control, Vol. 69, no.5, pp. 2060–2066.
(2023).
Random projection preserves stability with high probability.
JSIAM Letters, Vol. 15, pp. 17–20.