I am an Assistant Professor (Maître de Conférences) in the Data Science Department at EURECOM in France. My research focuses on statistics, machine learning, and applied mathematics, with applications in science and engineering.
Email: motonobu.kanagawa@gmail.com (preferred), motonobu.kanagawa@eurecom.fr (institutional)
Publications: Google Scholar
Research Topics
My research covers both theoretical and applied aspects of statistics, machine learning, and numerical computation. My main interests include:
Gaussian processes and kernel methods: Gaussian processes, reproducing kernel Hilbert spaces, and kernel methods for statistics, machine learning, and numerical computation (see our forthcoming book from Cambridge University Press).
Mathematical foundations: Convergence analysis, approximation theory, and theoretical properties of statistical, machine-learning, and numerical methods.
Probabilistic numerics and uncertainty quantification: Numerical integration, interpolation, Bayesian and kernel quadrature, and sequential and adaptive experimental design.
Nonparametric statistics and machine learning: Distribution comparison, variable selection, nearest-neighbour methods, missing-data imputation, and related statistical and machine-learning methods.
My research is also motivated by practical problems arising in science and engineering.
Industrial collaborations
I collaborate with companies, research institutes, and public-sector organizations on problems involving statistical and mathematical modelling, data analysis, machine-learning systems, simulation-based analysis, and data-driven decision support.
My approach is to work closely with collaborators to understand the practical problem, and then develop the simplest solution that is robust and easy to maintain. As a methodological researcher, I value such collaborations because they expose me to concrete real-world problems and constraints that can motivate new statistical and mathematical questions.
For example, I have supervised funded PhD projects with SAP Labs France on missing-data imputation and uncertainty quantification for solar-power forecasting, and with the National Institute for Environmental Studies (Japan) on machine learning for satellite-based greenhouse-gas retrieval and cloud screening. I have also supervised projects with research engineers funded by the Government of Monaco on storm-surge and tsunami simulation. At EURECOM, I have worked on timetable scheduling using mixed-integer linear programming.
I am currently involved in the ANR GUNESROSES project, collaborating with EDF (Électricité de France), CEA (French Alternative Energies and Atomic Energy Commission), ASNR (French Authority for Nuclear Safety and Radiation Protection), and other research partners on uncertainty quantification for nuclear-safety analysis.
If you have a problem related to data analysis, statistical/mathematical modelling, or machine learning that you would like to discuss, please feel free to contact me by email. I would be happy to exchange ideas by email or arrange a meeting to discuss the problem and possible approaches.
There are several possible forms of collaboration. For exploratory work, an EURECOM semester project allows a student to work on a problem for about three months without requiring research funding from the company or external organization. More substantial projects can be developed through internships, industry-linked PhD research, projects employing research engineers, or longer-term joint research.