Kansei Ushiyama

Ph.D. student,
Mathematical Informatics 3rd Laboratory,
Department of Mathemathical Informatics,
Graduate School of Information Science and Technology,
The University of Tokyo
Email: ushiyama-kansei074 (at) g.ecc.u-tokyo.ac.jp

Reserch Interests

Education

Mar. 2023
Master of Information Science and Technology, The University of Tokyo
(Supervisor: Prof. Takayasu Matsuo)
Thesis: Exploiting Numerical Analytical Concepts for Continuous Optimization: Numerical Stability and Discrete Chain Rule

Mar. 2021
Bachelor of Engineering, The University of Tokyo
(Supervisor: Prof. Yuki Izumida)
Thesis: A Study on the Mechanism of Synchronization Promotion by Uncorrelated Noise in Coupled Oscillator Systems (in Japanese)

Mar. 2017
Nagano prefectural Nagano senior high school.

Preprints

  1. Deriving Optimal Rates of Continuous-time Accelerated First-order Methods via Performance Estimation Problems
    [METR]
    K. Ushiyama, S. Sato, T. Matsuo, preprint.

Refereed Journal Articles and Conference Proceedings

  1. Differentiable Pareto-Smoothed Weighting for High-Dimensional Heterogeneous Treatment Effect Estimation
    Y. Chikahara, K. Ushiyama, to appear in Proceedings of the Fortieth Conference on Uncertainty in Artificial Intelligence.

  2. Analysis of continuous dynamical system models with Hessians derived from optimization methods
    T. Kamijima, S. Sato, K. Ushiyama, T. Matsuo, K. Tanaka, to appear in JSIAM Letters.

  3. Properties and practicability of convergence-guaranteed optimization methods derived from weak discrete gradients
    [journal]
    K. Ushiyama, S. Sato, T. Matsuo, to appear in Numer. Algorithms.

  4. A unified discretization framework for differential equation approach with Lyapunov arguments for convex optimization
    [arXiv][proceeding]
    K. Ushiyama, S. Sato, T. Matsuo, NeurIPS 2023.

  5. Essential convergence rate of ordinary differential equations appearing in optimization
    [arXiv][journal]
    K. Ushiyama, S. Sato, T. Matsuo, JSIAM Letters, 14(2022), 119–122.

  6. Deriving efficient optimization methods based on stable explicit numerical methods
    [journal]
    K. Ushiyama, S. Sato, T. Matsuo, JSIAM Letters, 14(2022), 29–32.

Talks

  1. Convergence of optimisation methods and stability of numerical methods
    K. Ushiyama, T. Saegusa, S. Sato, T. Matsuo, Auckland Numerical Analysis WoRKshop, Auckland (New Zealand), March 25–29, 2024.

  2. Extending discrete gradients for unified description and analysis of optimization methods
    K. Ushiyama, S. Sato, T. Matsuo, ANODE 2023, Auckland (New Zealand), February 20–24, 2023.

Awards

Oct. 2024
Student Excellent Presentation Award, The 2024 Fall National Conference of Operations Research Society of Japan

Dec. 2023
Student Excellent Presentation Award, The 26th Information-based Induction Sciences Workshop

Jul. 2023
Best Paper Award in JSIAM Letters, Japan Society for Industrial and Applied Mathematics

Mar. 2023
Dean’s award from Graduate School of Information Science and Technology, The University of Tokyo

Jun. 2022
Best Presentation Award, Japan Society for Industrial and Applied Mathematics

Fellowships

Apr. 2024 – Mar. 2026
JSPS Research Fellowship for Young Scientists, DC2

Apr. 2023 – Mar. 2026
JST Support for Pioneering Research Initiated by the Next Generation (SPRING) Program, The University of Tokyo “Advanced Human Resource Development Leading Green Transformation (GX) (SPRING GX)” Project student