Researcher's Profile

Incorporation of domain knowledge into machine learning
Machine learning is a framework that can acquire useful patterns from real-world data, and its performance largely depends on the amount and the quality of training data. By incorporating domain knowledge, such as scientific theories and empirical laws, we may improve not only the performance but also the interpretability of machine learning. We study the methods to this end, such as direct combination of machine learning and scientific models and inverse problem solving for complex simulators using machine learning. These methods have also been applied to various scientific and industrial problems.

Data-driven dynamical systems
Dynamical systems are ubiquitous in physical and biological phenomena, and analysis and prediction of dynamical systems are fundamental problems in various discplines. We study data-driven methods for analyzing dynamical systems, such as operator-theoretic view based on the Koopman operator and dynamic mode decomposition.