Minseok Jeong

Hi, I am Minseok Jeong, a Ph.D. student at the KAIST ACSS Lab, advised by Prof. SooJean Han. I received my B.E. in System Management Engineering (Industrial Engineering) from SKKU and my M.S. in Electrical and Electronic Engineering from GIST, where I worked with Prof. Euiseok Hwang in the IIS Lab.

My research is guided by a central question:

Can we identify representations for tractable prediction and control of dynamical systems?

I approach this question by studying linear structure in representations, drawing on ideas from operator theory and statistical learning theory.

My current research explores three directions through this lens: Koopman theory for controlled system, the linear representation hypothesis for vision-language-action (VLA) models, and kernel method for novelty detection.

If any of these topics interest you, please feel free to reach out!

news

Jan 23, 2026	L4DC 2026 🎉 “Scalable Infinitesimal Generator–Based Koopman Learning for Long-Horizon Prediction”
Sep 18, 2025	NeurIPS 2025, Spotlight 🎉 “Mitigating Instability in High Residual Adaptive Sampling for PINNs via Langevin Dynamics” (688 out of 21,575 submissions, top 3.2%)

selected publications

In press

Scalable Infinitesimal Generator–Based Koopman Learning for Long-Horizon Prediction

Minseok Jeong, SooJean Han, and Hyosang Shin

In Proceedings of the Learning for Dynamics and Control Conference (L4DC), 2026

Abs

Koopman operator theory provides a linear representation of nonlinear dynamics, offering strong potential for long-horizon prediction. However, most existing methods rely on snapshot-based one-step fitting and often suffer from error accumulation over extended rollouts. To address this issue, prior efforts have explored extending the training horizon or developing model-based approaches that leverage physics-informed formulations—such as infinitesimal generator fitting—to better exploit the underlying system dynamics. In this paper, we show that local transition fitting itself fundamentally limits long-horizon stability, whereas generator-based learning provides a more principled alternative. Despite their reliable advantages, generator-based learning schemes remain computationally demanding, as training dense MLP-based dictionaries with Jacobian computations is often intractable in practice. Building upon this line of work, we introduce a Random Fourier Feature–lifted Koopman formulation (RFF-PIKN) that employs shift-invariant RKHS representations for scalable and stable approximation, reducing computational cost under repeated resampling. We theoretically show that RFF-PIKN converges to a local minimizer of the true population generator risk, enabling accurate and stable long-horizon prediction. These results highlight the promise of RKHS-based physics-informed Koopman learning in advancing model-based prediction frameworks.
Published

Mitigating Instability in High Residual Adaptive Sampling for PINNs via Langevin Dynamics

Minseok Jeong^*, Giup Seo^*, and Euiseok Hwang

In Proceedings of the Conference on Neural Information Processing Systems (NeurIPS), 2025

Spotlight

Abs

Recently, physics-informed neural networks (PINNs) have gained attention in the scientific community for their potential to solve partial differential equations (PDEs). However, they face challenges related to resource efficiency and slow convergence. Adaptive sampling methods, which prioritize collocation points with high residuals, improve both efficiency and accuracy. However, these methods often neglect points with medium or low residuals, which can affect stability as the complexity of the model increases. In this paper, we investigate this limitation and show that high residual-based approaches require stricter learning rate bounds to ensure stability. To address this, we propose a Langevin dynamics-based Adaptive Sampling (LAS) framework that is robust to various learning rates and model complexities. Our experiments demonstrate that the proposed method outperforms existing approaches in terms of relative error, and stability across a range of enviroments, including high-dimensional PDEs where Monte Carlo integration-based methods typically suffer from instability.