JHIN SHEO YON (Layla)

Ph.D. Student, School of Computing, KAIST
Big Data Analytics and Learning Lab (BigDyL)
Advisor: Prof. Noseong Park

I am a Ph.D. student in the School of Computing at KAIST, advised by Prof. Noseong Park in the Big Data Analytics and Learning Lab (BigDyL).

My research focuses on time-series forecasting and anomaly detection, with an emphasis on deep learning methods grounded in differential equations and operator learning. I am particularly interested in time-series foundation models and probabilistic scenario forecasting to support decision-making under uncertainty.

E-mail : sheoyon.jhin@kaist.ac.kr 

Research Advisor : Noseong Park

Ph.D student @BigDyL

IT ALL BEGINS WITH AN IDEA

✦

IT ALL BEGINS WITH AN IDEA ✦

  • KDD 2021 - ACE-NODE

  • WWW 2022 - EXIT

  • AAAI 2023 - LEAP

  • KDD 2023 - PAD

  • KDD 2024 - CONTIME

  • ICLR 2026 - FOLD

News

  • “TimeOperator: A Function-to-Function Approach to Time Series Modeling“, Conference on Neural Information Processing Systems (NeurIPS), 2026, Sheo yon Jhin, B. Aditya Prakash, and Noseong Park

  • "HyBrain: Spatiotemporal Hyperedges for EEG Seizure Detection and Prediction," ACM International Conference on Information and Knowledge Management (CIKM), 2026, Hyunju Kim, Sheo Yon Jhin, Nosing Park, Nabil Imam

  • Awarded the Korean Presidential Science Scholarship (Ph.D., 2026) for outstanding academic achievement.

  • “Free Point-wise Anomaly Detection via Fold-bifurcation“, ICLR 2026, Sheo Yon Jhin, Noseong Park

  • “HINTS: Extraction of Human Insights from Time-Series without External Sources”, AAAI 2026 (AI4TS workshop) , Sheo Yon Jhin, Noseong Park

  • I will visit Prof. Aditya Prakash in Georgia Tech for the collaboration of designing time series foundation model.

  • Awarded the Korean Presidential Science Scholarship (Ph.D., 2025) for outstanding academic achievement.

  • "Addressing Prediction Delays in Time Series Forecasting: A Continuous GRU Approach with Derivative Regularization", KDD 2024, Sheo Yon Jhin, Seojin Kim, and Noseong Park

  • "Attentive Neural Controlled Differential Equations for Time-series Classification and Forecasting", KAIS Journal 2024( IF=3.205 ), Sheo Yon Jhin, Heejoo Shin, Sujie Kim, Seoyoung Hong, Solhee Park, Noseong Park, Seungbeom Lee, Hwiyoung Maeng, and Seungmin Jeon  

  • "Precursor-of-Anomaly Detection for å Time Series", KDD 2023, SheoYon Jhin, Jaehoon Lee, Noseong Park

  • "Learnable Path in Neural Controlled Differential Equations", AAAI 2023, Sheo Yon Jhin, Minju Jo, Seungji Kook, Noseong Park

  • "LORD: Lower-Dimensional Embedding of Log-Signature in Neural Rough Differential Equations", ICLR 2022,  Jaehoon Lee, Jinsung Jeon, Sheo yon Jhin, Jihyeon Hyeong, Jayoung Kim, Minju Jo, Seungji Kook, and Noseong Park

  • "EXIT: Extrapolation and Interpolation-based Neural Controlled Differential Equations for Time-series Classification and Forecasting", WWW 2022, Sheo Yon Jhin, Jaehoon Lee, Minju Jo, Seungji Kook, Jinsung Jeon, Jihyeon Hyeong, Jayoung Kim and Noseong Park

  • "Attentive Neural Controlled Differential Equations for Time-series Classification and Forecasting", ICDM 2021, Sheo Yon Jhin, Heejoo Shin, Seoyoung Hong, Solhee Park, and Noseong Park

  • "ACE-NODE: Attentive Co-Evolving Neural Ordinary Differential Equations", KDD 2021, Sheo Yon Jhin, Minju Jo, Taeyong Kong, Jinsung Jeong, and Noseong Park

  • "DPM: A Novel Training Method for Physics-Informed Neural Networks in Extrapolation", AAAI 2021, Jungeun kim, Kookjin Lee, Dongeun Lee, Sheo Yon Jhin, and Noseong Park

Publications

Research Programs

  • Developing continuous-time neural architectures based on differential equations (ODE,CDE,RDE) for irregular time-series modeling, representation learning, forecasting, and classification.

  • Developing reusable function-to-function foundation models that represent time series as continuous functions rather than fixed-grid sequences, enabling forecastingm imputationm and classification through operator learning.

  • Improving forecasting by learning latent structural and behavioral dynamics, including decomposition-based learning, human-factor modeling, and volatility-aware temporal representations.

  • Modeling abnormal temporal transitions through dynamical systems for early-warning signals and precise point-wise anomaly detection.

  • Probabilistic scenario forecasting, woprld models, and causal prior learning for decision-making under uncertainity.

CV

Experiences

Visiting Researcher (Jan. 2026 ~ July. 2026)

Georgia Institute of Technology (Georgia Tech), Atlanta, GA, United States

Host: Prof. B. Aditya Prakash

  • Time-series foundation models (Operator learning, zero-shot forecasting)

  • Multi-modal learning (time-series and text generation)

  • Scenario Forecasting

  • Protein Analysis with Molecular Dynamics

Visiting Researcher (Sep. 2024 ~ Feb. 2025)

Electronics and Telecommunications Research Institute (ETRI), Daejeon, South Korea

  • Led a project to forecast operating profit for major KOSPI-listed firms using financial statement time-series data.

  • Conducted feature selection and engineering using causal inference, fine-tuned SOTA models for industrial data.

  • Proposed a hybrid forecasting model combining efficiency of DLinear with accuracy of neural CDEs.

IT ALL BEGINS WITH AN IDEA

✦

IT ALL BEGINS WITH AN IDEA ✦

STILL HAVE QUESTIONS? let’s chat.