Jin-Young Jung, Integrated Ph.D. Course
Research interest
Jin-Young Jung is a Ph.D. student at Kyungpook National University, advised by Prof. Hyun-Chul Kim. His research focuses on developing real-time adaptive learning systems by establishing mathematical connections between biological principles and modern deep learning architectures like Transformers (ViT, Perceiver). Key research areas include:
The core problem: Online Continual Learning & Neural Dynamics
Theoretical analysis: Catastrophic Forgetting & Gradient Confusion & Sharpness
Proposed approach: Sparse Activations and Updates
Learning Algorithm: Iterative Attention and Predictive Coding
Education
Mar. 2022 - Present: Integrated M.S. and Ph.D., Dept. of Artificial Intelligence, Kyungpook National University, Daegu, Republic of Korea
Mar. 2016 - Feb. 2020: B.S., Electronics Engineering, Kyungpook National University, Daegu, Republic of Korea
Work experience
Oct. 2021 - Mar. 2022 : Collaborative Research Intern, Transaction Analysis, OKX (via Kyunpook National University)
Award
Best paper award, Brain Engineering Society of Korea (BESK), 2022 Summer Workshop
Best paper award, KNU-EE-Research Congress (KNU-EERC), 2025
Best poster award, Brain Engineering Society of Korea (BESK), 2022 Summer Workshop
Best poster award, Korean Society for Human Brain Mapping (KHBM) 2022 Fall Workshop
Tutorial, invited talk, and seminar
Jul. 10, 2025, Unified Gradient Conflict Control for Robust, Brain-Like Learning under Shift, Brain Engineering Society of Korea (BESK) 2025 Summer Workshop (Oral presentation, Best paper award)
Jul. 1, 2025, Unified Gradient Conflict Control: Architectural Sparsity and Gradient Regularization for Non-I.I.D. Adaptation, KNU-EE Research Congress (KNU-EERC) 2025 (Oral presentation, Best paper award)
Sep. 27, 2024, Sparceiver: Leveraging Dynamic Sparsity in Transformers for Better Shift Adaptation and Interpretability, KPK Workshop, Daegu, Republic of Korea
Sep. 26, 2024, Sparceiver: A Sparse Transformer for Mitigating Catastrophic Forgetting through Reducing Gradient Confusion, Brain Image Research Seminar (online)
May 17, 2024, Sparceiver: Overcoming Gradient Confusion with Dynamic Attention Threshold in Transformers, Korean Society for Industrial Applied Mathematics (KSIAM) Spring Conference, Daegu, Republic of Korea
June 29, 2023, Exploring Brain-plausibility and Modifications for the Perceiver, Brain Image Research Seminar (online)