Naoya Takeishi is a researcher at the University of Tokyo. His interests include machine learning for scientific problems, particularly the integration of machine learning with scientific models, and the data-driven analysis of dynamical systems.
Contact
ntake[at]g.ecc.u-tokyo.ac.jp
I no longer monitor the inboxes of my old addresses (…@ailab.t.u-tokyo.ac.jp, …@hesge.ch, …@riken.jp).
Emails from my current address (…@g.ecc.u-tokyo.ac.jp) are sometimes marked as spam automatically.
If you don’t receive a reply for a while, please check your spam folder (it’s also possible I’m simply away).
Recent activities
| 2025-03-01 | Invited talk on hybrid modeling @ the Workshop on Functional Inference and Machine Intelligence [slides] |
|---|---|
| 2024-12-14 | I will be a panelist @ the workshop on Machine Learning and the Physical Sciences, NeurIPS 2024 |
| 2024-10-22 | Invited talk on ML and scientific models @ The 2024 Fall Meeting of the Seismological Society of Japan [slides] |
| 2024-09-23 | I joined the Editorial Board of the journal Machine Learning: Science and Technology (MLST) |
| 2024-07-15 | Invited talk on neural nets and Koopman operator learning @ the workshop on Koopman Operators in Robotics, RSS 2024 [slides] |
| 2024-05-29 | Tutorial talk on ML and scientific models @ The 38th Annual Conference of the Japanese Society for Artificial Intelligence [slides] |
| 2023-08-03 | Invited talk on ML and scientific models @ The 46th Annual Meeting of the Japan Neuroscience Society |
| 2023-07-28 | Co-organized the workshop on Synergy of Scientific and ML Modeling @ ICML 2023 |
Research interests
ML for Science: Machine learning and scientific models
Effective integration of domain-specific knowledge, such as scientific models and simulators, into statistical machine learning. For example, we developed a deep dynamics model with prior knowledge of stable invariant sets. We formulated a method for learning dynamical systems with side information. We proposed a regularized learning method for properly learning a deep grey-box models for robust and physically-interpretable generative modeling. We further discussed the methods for learning such hybrid models.
I am also interested in machine learning-based solution of forward problems (e.g., solving differential equations) and inverse problems (e.g., inference of simulation parameters). For example, we studied a method for reliable neural simulation-based inference and a way to incorporate multifidelity information in simulation-based inference.
Data-driven dynamical systems
Analysis of dynamical systems using data-driven methods, such as dynamic mode decomposition (DMD) and its variants; as well as their underlying theory based on the Koopman operator. We proposed, for example, Bayesian DMD, DMD for random dynamics, DMD with neural net observables, nonnegative DMD, time-varying DMD, and discriminant DMD for labeled time-series. We also introduced the use of multiple kernel learning for diffusion-map-based Koopman analysis.