Keita Kume, Doctor (Engineering)
Assistant Professor, Dept. of Information and Communications Engineering, Institute of Science Tokyo
- Member of Prof. Isao Yamada's laboratory
- kume [at] sp.ict.e.titech.ac.jp
- Google Scholar Citations
- Research map
- T2R2 Tokyo Tech Research Repository
- Japanese Page
About me
Research Interests
- Signal processing, mechine learning, inverse problem
- Optimization (manifold optimization, nonsmooth optimization)
Lectures
- 2025: Basic Computer Programming (Institute of Science Tokyo)
- 2024: Basic Computer Programming (Tokyo Institute of Technology)
- 2019-2023(TA): Signal and System Analysis (Tokyo Institute of Technology)
Education
- March, 2024: Doctor's degree (Engineering), Tokyo Institute of Technology (supervisor: Prof. Isao Yamada)
- March, 2021: Master's degree (Engineering), Tokyo Institute of Technology (supervisor: Prof. Isao Yamada)
- March, 2019: Bachelor's degree (Engineering), Tokyo Institute of Technology (supervisor: Prof. Isao Yamada)
Work Experience
- October, 2024-Present: Assistant Professor, Dept. of Information and Communications Engineering,
Institute of Science Tokyo
(Institute of Science Tokyo opened on October 1, 2024, following the merger between Tokyo Institute of Technology and Tokyo Medical and Dental University) - April, 2024-September, 2024: Assistant Professor, Dept. of Information and Communications Engineering, Tokyo Institute of Technology
- April, 2021-March, 2024: Research Fellow (DC1) of Japan Society for the Promotion of Science (JSPS)
Award
- June, 2026: JSIAM 22th Young Researcher Presentation Award
- December, 2024: IEEE SPS Japan Student Conference Paper Award
- March, 2021: Outstanding master's student award (Tokyo Institute of Technology, Dept. of Information and Communications Engineering)
- November, 2021: IEICE SIP Symposium, Young Researcher Encouragement Award
News
-
A paper on a gradient mapping-type stationarity measure for the sum of two nonconvex nonsmooth functions and its applications to proximal gradient-type methods is available as a preprint paper. The paper provides an affirmative answer to an open question raised by [Olikier-Waldspurger, SIAM J. Optim., 2025] on whether every cluster point of a sequence generated by a proximal gradient method is a proximal stationary point under local Lipschitz smoothness of one component of the cost function. In addition, the paper extends our proximal variable smoothing algorithm to more general setting that the cost function is given by the sum of a locally Lipschitz continuous and lower regular function and a prox-regular function. The paper is an extended version of a paper presented at IEEE ICASSP2026 (2026/9/4)．
-
A paper on a DC composite optimization algorithm via variable smoothing for new robust phase retrieval method with nonconvex robust loss is available as a preprint paper . The paper proposes an inner-loop-free algorithm for minimization of the composition of a DC function and a smooth mapping by extending a variable smoothing algorithm for minimization of the composition of a weakly convex function and a smooth mapping. The paper is an extended version of a paper presented at EUSIPCO 2025 (2026/4/14).
-
The following two papers have been accepted for presentation at IEEE ICASSP2026@Barcelona (2026/1/21).
-
K. Kume and I. Yamada, "Minimization of Nonsmooth Weakly Convex Function over Prox-regular Set for Robust Low-rank Matrix Recovery," arXiv:2509.17549 (to appear in IEEE ICASSP 2026). (arXiv)
-
-
A paper on (i) a proposal of a hierarchical Nash equilibrium problem, i.e., a problem for finding a variational equilibrium over the solution set of a variational equilibrium problem, and (ii) its algorithm based on a fixed-point expression has been accepted for publication in Journal of Optimization Theory and Applications. The paper is an extended version of a paper presented at IEEE ICASSP 2025 (2025/12/25).
-
A paper on a single-loop optimization algorithm for the minimization problem of the composition of a weakly convex function and a differentiable mapping over a prox-regular set is available as a preprint paper (2025/10/22).
-
A paper on a nonconvex regularized convex model with non-quadratic data-fidelity function and its iterative algorithm based on fixed-point theory is available as a preprint paper . The paper is an extended version of a paper presented at EUSIPCO 2025 (2025/9/18).
-
Talked our recent study based on a preprint in EUROPT 2025 (Presentation slide) (2025/7/5)．
-
A paper on a single-loop optimization algorithm for the minimization problem of the sum of (i) L-smooth function, (ii) composition of a Lipschitz continuous weakly convex function and differentiable mapping, and (iii) proper lower semicontinuous convex function is available as a preprint paper. The paper is an extended version of a paper presented at IEEE ICASSP 2025 (2025/6/10).
-
Our two papers on (i) a nonconvex regularized convex model with non-quadratic data-fidelity function and (ii) a variable smoothing for minimization of the composite function of a difference-of-(weakly)convex function and a differentiable mapping have been accepted for presentation at EUSIPCO2025@Sicily (2025/5/21).
-
I'm going to talk about the following our study at EUROPT 2025@Southampton (29 Jun. - 2 Jul.). Many thanks to Dr. Alireza Kabgani (University of Antwerp) for the kind invitation to his organized session ! (2025/4/14)
-
K. Kume and I. Yamada, "A Proximal Variable Smoothing for Nonsmooth Minimization of the Sum of Three Functions Including Weakly Convex Composite Function," The 22nd EUROPT Conference on Advances in Continuous Optimization (EUROPT 2025) (Session: ''Smoothing techniques for nonsmooth optimization'' organized by Dr. Alireza Kabgani). (EUROPT 2025)
-
-
A paper on (i) a proposal of a hierarchical Nash equilibrium problem, i.e., a problem for finding a variational equilibrium over the solution set of a variational equilibrium problem, and (ii) its algorithm based on a fixed-point expression is available asa preprint paper. The paper is an extended version of a paper presented at IEEE ICASSP 2025 (2025/4/8).
-
A paper on a variable smoothing for minimization of the composite function of a difference-of-(weakly)convex function and a differentiable mapping is available as a preprint paper (2025/4/8).
-
A paper on a nonconvex regularized convex model with non-quadratic data-fidelity function is available asa preprint paper (2025/4/8).
-
A paper on design of a regularizer of designated isolated minimizers and its application to discrete-valued signal estimation problem is available as a preprint paper. The paper is an extended version of a paper presented at APSIPA ASC 2024 (2025/3/14).
-
Talked our recent study in Research Association of Optimization Theory and Algorithms (RAOTA) (Presentation slide in Japanese)
-
久米啓太, "弱凸関数と可微分写像からなる合成関数の最適化のための可変平滑化法と信号処理応用," 最適化の理論とアルゴリズム（RAOTA）第8回研究会. (RAOTA, Presentation slide in Japanese)
-
-
Recieved IEEE SPS Japan Student Conference Paper Award for the following IEEE ICASSP2024 paper (2025/1/1).
- K. Kume and I. Yamada, "A variable smoothing for nonconvexly constrained nonsmooth optimization with application to sparse spectral clustering," 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 9296-9300, 2024 [IEEE SPS Japan Student Conference Paper Award]. (poster) (official access, arXiv, poster)
-
The following two papers have been accepted for presentation at IEEE ICASSP2025@Hyderabad (2025/1/1).
- K. Kume and I. Yamada, "A Proximal Variable Smoothing for Nonsmooth Minimization Involving Weakly Convex Composite with MIMO Application," arXiv:2409.10934 (to appear in IEEE ICASSP2025). (arXiv)
- S. Matsuo, K. Kume and I. Yamada, "Hierarchical Nash Equilibrium over Variational Equilibria via Fixed-point Set Expression of Quasi-nonexpansive Operator," arXiv:2409.11094 (to appear in IEEE ICASSP2025). (arXiv)
-
A paper on nonconvexly constrained nonsmooth minimization of a weakly convex composite function is available as a preprint paper. The paper is an extended version of a paper presented at IEEE ICASSP 2024 (2024/12/12).
-
A paper entitled "A Discrete-Valued Signal Estimation by Nonconvex Enhancement of SOAV with cLiGME Model" has been accepted at APSIPA@Macau (2024/12/12).
-
A paper on nonsmooth minimization of sum of a weakly convex composite function and a proper lower semicontinuous convex function is available as a preprint paper (2024/9/19).
-
A paper on finding a hierarchical Nash equilibrium via its fixed-point expression is available as a preprint paper (2024/9/19).
2026
2025
2024
Publication
Preprint
Journal Paper (peer-reviewd)
International Conference (peer-reviewed)
Invited talk
Documents
弱凸関数と可微分写像からなる合成関数の最適化のための可変平滑化法と信号処理応用 (Presentation slide for the talk in RAOTA)