Sangjun Han (한상준)

I am currently a researcher at LG AI Research, developing AI models capable of predicting and generating unseen future states. My expertise spans across working with sequential data in diverse fields, from artistic to industrial applications (music generation, bio-signal processing, battery, time-series modelling, causal discovery). Now, I'm working on the time-series forecasting, leveraging LLM to enhance multi-modality. (Last updated: Feb 2025)

Publications

Referred Journals

• Moonyoung Kwon, Sangjun Han, Kiwoong Kim, Sung Chan Jun, "Super-Resolution for Improving EEG Spatial Resolution using Deep Convolutional Neural Network - Feasibility Study", Sensors, 2019 [paper]

Referred Conferences

• Sangjun Han, Hyeongrae Ihm, Moontae Lee, Woohyung Lim, "Symbolic Music Loop Generation with Neural Discrete Representations", Proc. of the 22nd International Society for Music Information Retrieval Conference (ISMIR), 2022 [paper] [code] [demo]

• Sangjun Han, Mooyoung Kwon, Sunghan Lee, Sung Chan Jun, "Feasibility of EEG Super-Resolution using Deep Convolutional Networks", IEEE International Conference on Systems, Man, and Cybernetics (SMC), 2018 (oral) [paper] [slide]

• Sunghan Lee, Sangjun Han, Sung Chan Jun, “EEG Hyperscanning for Eight or more Persons - Feasibility Study for Emotion Recognition using Deep Learning Technique”, Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA), 2018 [paper]

• Sangjun Han, Moonyoung Kwon, Sung Chan Jun, “EEG Spatial Super-Resolution using Deep Convolutional Linear Networks: a Simulation Study”, Korean Society of Medical & Biological Engineering (KOSOMBE), 2017 (best paper)

• Jinyoung Choi, Sangjun Han, Moonyoung Kwon, Hyeon Seo, Sehyeon Jang, Sung Chan Jun, “Study on Subject-Specific Parameters in Sleep Spindle Detection Algorithm”, International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2017 [paper]

• Jinyoung Choi, Sangjun Han, Kyungho Won, Sung Chan Jun, "Effect of Acoustic Stimulation after Sleep Spindle Activity", Sleep Medicine, 2017 [paper]

Workshop Papers

• Chanhui Lee, Juhyeon Kim, Yongjun Jeong, Yeom Yoon Seok, Juhyun Lyu, Jung-Hee Kim, Sangmin Lee, Sangjun Han, Hyeokjun Choe, Soyeon Park, Woohyung Lim, Kyunghoon Bae, Sungbin Lim, Sanghak Lee, "On Incorporating Prior Knowledge Extracted from Pre-trained Language Models into Causal Discovery", NeurIPS Workshop on Causality and Large Models, 2024 [paper]

• Suhee Yoon, Sanghyu Yoon, Hankook Lee, Sangjun Han, Ye Seul Sim, Kyungeun Lee, Hyeseung Cho, Woohyung Lim, "Diffusion-based Semantic-discrepant Outlier Generation for Out-of-Distribution Detection", NeurIPS Workshop on SyntheticData4ML, 2023 [paper]

• Sangjun Han, Hyeongrae Ihm, Woohyung Lim, "The Interface for Symbolic Music Loop Generation Conditioned on Musical Metadata", NeurIPS Workshop on Machine Learning for Creativity and Design, 2023 [paper] [code]

• Sangjun Han, Hyeongrae Ihm, DaeHan Ahn, Woohyung Lim, "Instrument Separation of Symbolic Music by Explicitly Guided Diffusion Model", NeurIPS Workshop on Machine Learning for Creativity and Design, 2022 [paper] [code] [demo]

• Hyeongrae Ihm, Sangjun Han, Woohyung Lim, "Composer AI with Tap-to-pitch Generator", NeurIPS Workshop on Machine Learning for Creativity and Design, 2021 [paper]

Preprints in arXiv

• Sangjun Han, Jiwon Ham, Chaeeun Lee, Heejin Kim, Soojong Do, Sihyuk Yi, Jun Seo, Seoyoon Kim, Yountae Jung, Woohyung Lim, "Flexible Control in Symbolic Music Generation via Musical Metadata", arXiv:2409:07467, 2024 [paper] [demo]

• Chanhui Lee, Juhyeon Kim, Yongjun Jeong, Juhyun Lyu, Jung-Hee Kim, Sangmin Lee, Sangjun Han, Hyeokjun Choe, Soyeon Park, Woohyung Lim, Sungbin Lim, Sanghak Lee, "Can We Utilize Pre-trained Language Models within Causal Discovery Algorithms?", arXiv:2311.11212, 2023 [paper]

• Sangjun Han, Hyeongrae Ihm, Woohyung Lim, "Systemic Analysis of  Music Representations from BERT", preprint arXiv:2306.04628, 2023 [paper] [code]

• Sangjun Han, Hyeongrae Ihm, Woohyung Lim, "Symbolic Music Loop Generation with VQ-VAE", preprint arXiv:2111.07657, 2021 [paper]

Education

Gwangju Institute of Science and Technology

• M.S. in School of Electrical Engineering and Computer Science 

• Gwangju, South Korea, Aug 2016 - Aug 2018 

Handong Global University

• B.S. in School of Computer Science and Electrical Engineering

• Pohang, South Korea, Feb 2011 - Aug 2016

Work Experience

LG AI Research

• Data Intelligence Lab, Applied AI Research Lab

• Seoul, South Korea, Dec 2020 - Current

LG Sciencepark (CIC of LG Electronics)

• AI Solution Development Team, AI Machine Learning Team

• Seoul, South Korea, Apr 2019 - Dec 2020

LG CNS

• AI Tech Team, AI Solution Development Team

• Seoul, South Korea, Jul 2018 - Apr 2019

Projects

LLM based Time-series Forecasting and Editing for Commodities

• I'm developing an editing module that adjusts predicted values based on political, demand-supply, environmental issues using LLMs. It'll be integrated into the forecasting system "Futurecast" from LG AI Research, which forecasts the price of commodities. [demo]

• from LG AI Research, Jan 2025 - Current

Continuous Optimization based Causal Discovery for Lithium Price Forecasting

• I have extracted causal graphs based on data-driven optimization, which explain the change in lithium price to assist human decision-making. [image]

• with LG AI Research & LG Energy Solution, June 2023 - Dec 2023

(Research Project) Causal Discovery for Time-series using Large Language Model as Prior

• Prof. Sanghack Lee from Seoul National University and Prof. Sungbin Lim from Korea University [paper]

• from LG AI Research, Apr 2023 - Apr 2024

(Project Leader) Symbolic and Audio Music Generation

• We are eager to develop a music generation system that empowers individuals to discover and express their musical identity. We are actively addressing the challenge of multimodal generation, including text, audio, and MIDI. [blog1] [blog2] [news] [video]

• from LG AI Research, Feb 2021 - Dec 2024

Prediction of Battery Life Cycle and its Capacity

• I have assessed various prediction models of battery discharge capacity, including tree-based modes and neural network models. I have proposed a novel neural network model that outperforms baselines through improved regularization. Also, several XAI techniques are applied to explain the model's outcome. [video]

• With the domain knowledge of lithium batteries, I have explored prediction models of the battery life cycle, addressing data scarcity. [blog] [video]

• with LG Energy Solution, Jan 2020 - Dec 2021

Discovering New Chemical Molecules with Specific Target Properties using Deep Generative Models

• I have developed a Web interface for Chemical VAE that generates novel molecules in SMILES form. [concept]

• with LG Chem, Jan 2020 - July 2020

(Research Project) Deep Learning for Multiple Time-series Probabilistic Forecasting

• Prof. Merve Bodur from University of Toronto [certificate]

• from LG Sciencepark, Jan 2020 - Dec 2020

Compressing Deep Learning Model for Vision Inspection

• I have tested multiple techniques of model compression in both CPU and GPU environments, achieving 0.03% model size and 0.11% latency on CPU without compromising the performance for vision inspection. They include efficient models, pruning, compiling, TensorRT, and Intel-MKL. [blog]

• with LG Display, Jan 2019 - Apr 2019

Object Detection based Automatic Payment System in Convenience Store

• I have experienced the whole process of AI systems, including data collection, labeling, model development, and system deployment. Also, I have explored and developed the classification component of object detection to enhance its robustness across diverse environmental conditions. [video]

• with GS Retails, July 2018 - Apr 2019

Patents

Simultaneous EEG Measurement Apparatus and Method for at least Two People

• South Korea, 10-2020-0017132

Teaching

Project Mentor

• (LINC Industry-Academia Collaboration) Introduction to AI Projects, Handong Global University, 2024

• (LG Academy) Predicting Remaining Useful Life for Display Predictive Maintenance, LG Display, 2020

• (LG Academy) PCA/Autoencoder based Predictive Maintenance for Camera Design Process, LG Innotek, 2020

• (LG Academy) Anomaly Detection by Affinity Propagation based Clustering for Battery Faults, LG Chem, 2020

• (LG Academy) Learning to Design Circuit by Reinforcement Learning (Soft Actor-Critic), LG Display, 2019

Teaching Assistant

• Database Systems, Handong Global University, 2016

• Introduction to Engineering Design, Handong Global University, 2016 [slide]

Presentations

Invited Talks, Tutorials, and Posters

• "Thoughts on Researching and Developing Artificial Intelligence", Dongseo University, 2024 [slide]

• "Recent Trends and Industrial Applications in Deep Learning", at Handong Global University, 2024 [slide]

• "Recent Trends and Industrial Applications in Deep Learning", at Korea Institute of Energy Technology, 2024 [slide]

• "AI's Impact on Industries: Bridging the Gap between Research and Application", at Dongseo University, 2023 [slide]

• "AI Music Composer, Bridging the Gap between Emotion and Technology", at Seoul/Busan LG Discovery Lab, 2023 [slide]

• "Linear Models: Anomaly Detection, Forecasting, and Causal Discovery", at Korea Institute of Energy Technology, 2023 [slide] [exercise]

• "Time-series Analysis and Forecasting for Energy Systems", at The Korean Society for New and Renewable Energy, 2023 [slide] [exercise] 

• "Applied AI Research for Industry (Representation, Forecasting, Generation)", at Kyungpook National University, 2023 [slide]

• "Why is Data Important for Business?", at Korea Institute of Energy Technology, 2022 [slide] [exercise]