RIKEN Center for Advanced Intelligence Project High-Dimensional Causal Analysis Team
Team Leader: Masaaki Imaizumi (Ph.D.)
Research Summary
The High-Dimensional Causal Analysis Team studies the causal structure in high-dimensional data. In today's world of advanced data acquisition, storage, and analysis technologies, not only the structures possessed by data, but also the data science techniques for analyzing them have become highly complex. This team aims to understand both the large-degree-of-freedom, high-dimensional, and complex structures of data and the techniques for analyzing them, and to construct and extend theories based on these understandings. As specific methodologies, we study modern high-dimensional statistics and deep learning theory, and use them to develop methods of analysis and statistical inference for causal structures.
Main Research Fields
- Informatics
- Mathematical & Physical Sciences
Related Research Fields
- Principles of Informatics/Statistical science
- Principles of Informatics/Mathematical informatics
Keywords
- High-dimensional statistics
- Statistical causal inference
- Deep learning theory
- Statistical learning theory
Selected Publications
- 1.
R.Zhang, M.Imaizumi, B.Schölkopf, K.Muandet.
"Instrumental Variable Regression via Kernel Maximum Moment Loss"
Journal of Causal Inference, 11(1), (2023). - 2.
M.Imaizumi, J.Schmidt-Hieber.
"On Generalization Bounds for Deep Networks Based on Loss Surface Implicit Regularization"
IEEE Transaction on Information Theory, 69(2), (2023). - 3.
M.Imaizumi, K.Fukumizu.
"Advantage of Deep Neural Networks for Estimating Functions with Singularity on Hypersurface"
Journal of Machine Learning Research, 23(111), (2022). - 4.
M.Kato, M.Imaizumi, K.McAlinn, S.Yasui, H.Kakehi.
"Learning Causal Models from Conditional Moment Restrictions by Importance Weighting"
International Conference on Learning Representations (spotlight), (2022). - 5.
A.Sannai, M.Imaizumi, M.Kawano.
"Improved Generalization Bounds of Group Invariant / Equivariant Deep Networks via Quotient Feature Spaces"
PMLR: Uncertainty on Artificial Intelligence, (2021). - 6.
R.Nakada, M.Imaizumi.
"Adaptive Approximation and Generalization of Deep Neural Network with Intrinsic Dimensionality"
Journal of Machine Learning Research 21(174), (2020). - 7.
M.Imaizumi, K.Fukumizu.
"Deep Neural Networks Learn Non-Smooth Functions Effectively"
PMLR: Artificial Intelligence and Statistics, (2019). - 8.
M.Imaizumi, T.Maehara, Y.Yoshida.
"Statistically Efficient Estimation for Non-Smooth Probability Densities"
PMLR: Artificial Intelligence and Statistics, (2018). - 9.
M.Imaizumi, K.Kato.
"PCA-based estimation for functional linear regression with functional responses"
Journal of Multivariate Analysis 163, (2018). - 10.
M.Imaizumi, T.Maehara, K.Hayashi.
"On Tensor Train Rank Minimization: Statistical Efficiency and Scalable Algorithm"
Advances in Neural Information and Processing Systems 30, (2018).
Lab Members
Principal investigator
- Masaaki Imaizumi
- Team Leader
Core members
- Ryuichiro Hataya
- Postdoctoral Researcher
- Guillaume Braun
- Postdoctoral Researcher
- Hanna Tseran
- Postdoctoral Researcher
- Shogo Nakakita
- Visiting Scientist
- Kota Matsui
- Visiting Scientist
- Jacopo Peroni
- Student Trainee
- Eshant English
- Student Trainee
- Mana Sakai
- Research Part-time Worker II
Contact Information
Nihonbashi 1-chome Mitsui Building, 15th floor,
1-4-1 Nihonbashi, Chuo-ku, Tokyo
103-0027, Japan
Email: masaaki.imaizumi [at] riken.jp