Takaharu Yaguchi
Takaharu Yaguchi works on machine learning for physics, studying models that preserve physical laws when they learn or simulate motion. Yaguchi coauthored Deep Energy-Based Modeling of Discrete-Time Physics. The work was conducted outside RIKEN. It is part of his research on learning representations of physical systems. Yaguchi directs RIKEN’s Computational Physics Machine Learning Team. The team develops scientific learning methods that respect physical laws, including energy conservation, and studies their theoretical behavior.
Timeline
-
2020
Learning physical dynamics
Yaguchi coauthored Deep Energy-Based Modeling of Discrete-Time Physics. The work was conducted outside RIKEN. It forms part of his research on learning representations of physical systems.
-
2021
Physical structure and model gradients
Shared papers studied learning Hamiltonian equations through neural symplectic forms and computing exact gradients of neural ordinary differential equations with a symplectic adjoint method. The work connected numerical methods with trainable models of dynamics.
-
2025
Operators and coupled systems
Yaguchi coauthored research on energy-consistent neural operators for Hamiltonian and dissipative partial differential equations. Another paper used Poisson-Dirac neural networks to model coupled dynamical systems across domains. The latter work was conducted outside RIKEN.
-
As of October 2026
Reliable scientific learning
Yaguchi directs RIKEN’s Computational Physics Machine Learning Team. The team develops scientific learning methods that respect physical laws, including energy conservation, and studies their theoretical behavior.
Sources and credits
Photo: www.riken.jp; Takaharu Yaguchi profile photograph; 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.