Yuji Okamoto

Papers

Total Citations

H-Index

About

Yuji Okamoto’s research lies at the intersection of machine learning, dynamical systems, and control theory, with a focus on learning physically meaningful and stable dynamics from time-series data. His work addresses a fundamental challenge: ensuring that neural networks trained on observed data inherently respect key system properties like stability and dissipativity. In his most-cited study, “Learning Deep Dissipative Dynamics” (2025, 3 citations), Okamoto pioneers a method to guarantee dissipativity—a property that generalizes stability and input-output stability—in learned neural representations. This is critical for applications in robotics, physical modeling, and systems biology, where models must reliably interact with their environment. His earlier work, “Learning Deep Input-Output Stable Dynamics” (2022, 2 citations), tackled the specific problem of ensuring input-output stability, a cornerstone for systems that communicate with external forces. Though early in his career, Okamoto’s contributions are notable for bridging rigorous control-theoretic guarantees with modern deep learning, offering a principled path toward trustworthy, data-driven models. His research is poised to impact fields requiring robust, physically consistent simulations.