Kentaro Hoshisashi

I am a computational scientist whose work spans partial differential equations, statistical physics, and machine learning. I am an Honorary Research Fellow in the Department of Computer Science at University College London (UCL), and a quantitative researcher in EMEA Quantitative Research at SMBC Bank International in London. I completed my PhD in Computer Science at UCL in 2025.

My research uses physics-informed neural networks (PINNs) as a numerical method for the PDEs that govern stochastic systems, enforcing conservation laws, boundary behaviour, and the Fokker–Planck dynamics of probability densities directly in the loss. The core method, derivative-constrained PINNs for solving constrained PDEs, is the subject of my most recent paper in Phys. Rev. E. I apply it to probability-density-consistent calibration of stochastic local volatility models and to arbitrage-free deep calibration of implied volatility surfaces, implemented for performance and reproducibility in JAX. This work received the Best Paper Award at the IEEE Symposium on Computational Intelligence for Financial Engineering and Economics (CIFEr) in 2024.

I came to this from experimental physics. I hold an MSc from the University of Tokyo and a BSc from the Tokyo Institute of Technology, and my earlier work measured single-molecule dynamics using diffracted X-ray tracking and atto-newton X-ray radiation pressure, recovering picometre-scale motion from noisy, high-dimensional data. Fluid dynamics or a volatility surface, the problem is much the same: writing down the physics of a complex system and computing how it evolves.

Please use the links below for my code, publications, and profiles.