Masaaki Imaizumi
Masaaki Imaizumi studies the theory of learning from high-dimensional and complex data. His work includes a 2019 paper with Kenji Fukumizu on non-smooth functions and their 2022 JMLR analysis of deep networks for functions with singular structure. The latter establishes advantages under defined mathematical conditions. His 2023 papers address generalization bounds and high-dimensional contextual bandits. As of October 2026, Imaizumi directs RIKEN’s High-Dimensional Structure Theory Team.
- Now at
- RIKEN and University of Tokyo
- Works
- 94
- Citations
- 335
- h-index
- 10
Publication numbers from OpenAlex, as of 14 October 2026.
Timeline
-
2019
Learning non-smooth functions
Imaizumi and Kenji Fukumizu coauthored Deep Neural Networks Learn Non-Smooth Functions Effectively. The paper appeared at AISTATS in 2019. It is part of his research on how neural networks estimate functions that contain irregular structure.
-
2022
A mathematical advantage for depth
Their JMLR paper analyzes functions with singularities on hypersurfaces. It derives error rates and describes conditions where deep-network estimators outperform several other estimator classes.
-
2023
Generalization and decisions
His papers include work with Johannes Schmidt-Hieber on generalization bounds based on loss-surface implicit regularization. A separate shared study addresses contextual bandits in high dimensions without sparsity. Both examine the behavior of learning under specific mathematical conditions.
-
As of October 2026
High-dimensional theory at RIKEN
Imaizumi directs RIKEN’s High-Dimensional Structure Theory Team. The team uses mathematical statistics, probability and statistical mechanics. Its research topics include deep-learning theory, causal inference and the analysis of complex high-dimensional data.
Sources and credits
Photo: www.riken.jp; creator not stated. License: Not stated by source. Source: riken.jp. Cropped to a square and resized. Ask us to remove or replace this photo.