Sonoda Sho

ソノダショウ | Sonoda Sho

Affiliation and department：
Job title： Senior Scientist
Homepage URL (1)： https://sites.google.com/view/shosonoda/

Research field (1)： Intelligent informatics

Research keywords (12)： machine learning , deep learning , integral representation theory , ridgelet transform , neural network , representation theory , harmonic analysis , kernel quadrature , Wasserstein geometry , particle filtering , quantum machine learning , automated theorem prover

Research theme for competitive and other funds (4)：

2024 - 2028 Deepening the analysis of deep learning through functional space theory
2021 - 2025 複雑データに内在する深層構造の理論と応用
2018 - 2022 Transportation analysis of deep neural networks
2015 - 2017 積分表現理論によるディープニューラルネットの解析と設計指標の開発

Papers (26)：

園田翔. 深層ニューラルネットの構成的普遍近似定理-群表現論的方法-. 日本神経回路学会誌. 2024. 31. 4. 177-186
Sho Sonoda, Isao Ishikawa, Masahiro Ikeda. A unified Fourier slice method to derive ridgelet transform for a variety of depth-2 neural networks. Journal of Statistical Planning and Inference. 2024. 233. 106184
Toshinori Kitamura, Tadashi Kozuno, Masahiro Kato, Yuki Ichihara, Soichiro Nishimori, Akiyoshi Sannai, Sho Sonoda, Wataru Kumagai, Yutaka Matsuo. A Policy Gradient Primal-Dual Algorithm for Constrained MDPs with Uniform PAC Guarantees. First Reinforcement Learning Safety Workshop. 2024
園田翔. 深層学習と調和解析/写像の幅と深さを計算する. 数学セミナー2024年6月号. 2024. 63. 6. 74-79
Y. Hashimoto, S. Sonoda, I. Ishikawa, A. Nitanda, T. Suzuki. Koopman-Based Bound for Generalization: New Aspect of Neural Networks Regarding Nonlinear Noise Filtering. The Twelfth International Conference on Learning Representations. 2024

ｍore...

MISC (14)：

S. Sonoda, Y. Hashimoto, I. Ishikawa, M. Ikeda. Unified Universality Theorem for Deep and Shallow Joint-Group-Equivariant Machines. ArXiv preprint: 2405.13682v2. 2024
Hayata Yamasaki, Sho Sonoda. Exponential Error Convergence in Data Classification with Optimized Random Features: Acceleration by Quantum Machine Learning. 2022
S. Sonoda, I. Ishikawa, M. Ikeda. Ghosts in Neural Networks: Existence, Structure and Role of Infinite-Dimensional Null Space. 2021
Sho Sonoda. Fast Approximation and Estimation Bounds of Kernel Quadrature for Infinitely Wide Models. 2020
Sho Sonoda, Isao Ishikawa, Masahiro Ikeda, Kei Hagihara, Yoshihiro Sawano, Takuo Matsubara, Noboru Murata. The global optimum of shallow neural network is attained by ridgelet transform. 2018

ｍore...

Books (1)：

Data Science and Machine Learning: Mathematical and Statistical Methods
2022 ISBN:9784807920297

Lectures and oral presentations (72)：

Deep Ridgelet Transform: Harmonic Analysis for Deep Neural Network
(The 14th AIMS Conference Session on Understanding the Learning of Deep Networks: Expressivity, Optimization, and Generalization 2024)
Deep Ridgelet Transform: Harmonic Analysis for Deep Learning Machine
(Statistical Models and Mathematical Optimization Based on Geometric Structures 2024)
Deep Ridgelet Transform: Harmonic Analysis for Deep Neural Network
(One World Seminar Series on the Mathematics of Machine Learning 2024)
Deep Ridgelet Transform: Harmonic Analysis for Deep Neural Network
(Applied Geometry for Data Sciences 2024)
Deep Ridgelet Transform: Harmonic Analysis for Deep Neural Network
(Workshop on Mathematical Foundations of Machine Learning at CIRM 2024)

ｍore...

Professional career (1)：

博士(工学) (早稲田大学)

Awards (5)：

2023/10 - 第26回情報論的学習理論ワークショップ (IBIS2023) 優秀プレゼンテーション賞ファイナリストニューラルネットのパラメータに作用する双対群と普遍性定理の表現論的証明
2022/06 - 日本応用数理学会第18回若手優秀講演賞(2021年度) 積分表現ニューラルネットが定める積分方程式の一般解
2017/04 - 早稲田大学第2回WIRPワークショップ最優秀賞深層ニューラルネットの輸送解釈とWasserstein幾何学的解析
2016/11 - 第19回情報論的学習理論ワークショップ (IBIS2016) 学生最優秀プレゼンテーション賞無限層デノイジング・オートエンコーダーの輸送理論解釈
2012/03 - 日本鉄鋼協会計測・制御・システム研究賞物理・統計的モデリングによる取鍋内溶鋼温度の高度予測技術

※ Researcher’s information displayed in J-GLOBAL is based on the information registered in researchmap. For details, see here.

Return to Previous Page