Ryotaro Mitsuboshi's Home Page (EN/JP)

[LinkedIn] [GitHub] [ORCiD] [ResearchGate]

Ryotaro Mitsuboshi studies machine learning algorithms and theories.
His current interest is theoretically guaranteed boosting algorithms.
He uses Rust, C/C++, Python3, and Haskell for programming.

Research interests

• Machine Learning
• Boosting
• Support Vector Machines
• Statistical Learning Theory
• Online Learning
• Convex optimization
• Frank-Wolfe algorithms

Publications

1. Ryotaro Mitsuboshi, Kohei Hatano, and Eiji Takimoto.
   Boosting as Frank-Wolfe.
   Preprint
   [arXiv] [code]
2. Yuta Kurokawa, Ryotaro Mitsuboshi, Haruki Hamasaki, Kohei Hatano, Eiji Takimoto, and Holakou Rahmanian.
   Extended Formulations via Decision Diagrams.
   COCOON 2023
   [paper] [arXiv] [code] [slide]
3. Ryotaro Mitsuboshi, Kohei Hatano, and Eiji Takimoto.
   Solving Linear Regression with Insensitive Loss by Boosting.
   IEICE Transactions on Information and Systems 2024
   [paper] [code]
4. Ryotaro Mitsuboshi, Kohei Hatano, and Eiji Takimoto.
   Online Combinatorial Linear Optimization via a Frank-Wolfe-based Metarounding Algorithm.
   Preprint
   [arXiv] [code]

Projects

1. MiniBoosts
   A collection of boosting algorithms written in Rust.
2. Bandit
   A small collection of bandit algorithms written in Rust.