Yesom Park
Hedrick Assistant Adjunct Professor
Department of Mathematics, UCLA
Research Interests:
Scientific Machine Learning · Numerical Analysis
I am a Hedrick Assistant Adjunct Professor in the Department of Mathematics at UCLA, where I work with Stanley Osher and Hayden Schaeffer.
My research explores the mathematical foundations of modern computation at the intersection of differential equations, machine learning, and numerical analysis. Rather than pursuing incremental algorithmic improvements, I seek mathematical principles that lead to computational methods that are not only faster or more accurate, but fundamentally better formulated. I rethink computational problems from first principles by identifying appropriate mathematical representations, uncovering hidden structures, and establishing connections across seemingly distinct fields.
This perspective shapes my research in scientific machine learning, neural representations, generative modeling, optimal transport, 3D computer vision, and convergence and preconditioning analysis.
I received my Ph.D. in Mathematics from Seoul National University, advised by Myungjoo Kang.
selected publications
- SISCScalable Fixed-Point Framework for High-Dimensional Hamilton-Jacobi EquationsSIAM Scientific Computing, 2026
- IJCVNeural Shortest Path for Surface Reconstruction from Point CloudsInternational Journal of Computer Vision, 2026
- SINUMLocalized Estimation of Condition Numbers for MILU Preconditioners on a GraphSIAM Journal on Numerical Analysis, 2026
- ICLRNeural Hamilton–Jacobi Characteristic Flows for Optimal TransportIn International Conference on Learning Representations, 2026
- SISCNeural Implicit Solution Formula for Efficiently Solving Hamilton–Jacobi EquationsSIAM Journal on Scientific Computing, 2025
- NeurIPSHow does PDE order affect the convergence of PINNs?Advances in Neural Information Processing Systems, 2024
- JCPReSDF: Redistancing implicit surfaces using neural networksJournal of Computational Physics, 2024
- NeurIPSp -Poisson surface reconstruction in curl-free flow from point cloudsAdvances in Neural Information Processing Systems, 2023