Efficient Deep Learning Inference, Training, and Adaptation
Embedded Computer Vision, NLP (Natural Language Processing), Reinforcement Learning, Multitask Learning, and Meta Learning
On-Device Interpretable and Explainable AI (XAI)
On-Device Neural Architecture Search (On-Device NAS)
Generative AI on Embedded Systems (Image/Video Generation and Language Models)
On-Device Self-Supervised Learning and Continual Learning
Online Dataset Adaptation and Neural Architecture Search
Real-time Machine Learning
TransPL: VQ-Code Transition Matrices for Pseudo-Labeling of Time Series Unsupervised Domain Adaptation (ICML 2025)
PPT: Patch Order Do Matters In Time Series Pretext Task (ICLR 2025)
SMMF: Square-Matricized Momentum Factorization for Memory-Efficient Optimization (AAAI 2025)
Designing Extremely Memory-Efficient CNNs for On-device Vision Tasks (ACCV 2024)
Best Application Paper Award (Top 0.72%)
CAFO: Feature-Centric Explanation on Time Series Classification (KDD 2024)
Best Poster Award (2024 UNIST AI Technology Open Workshop)
On-Device Neural Architecture Search (SenSys 2023)
Softmax Output Approximation (NeurIPS 2023)
MicroDeblur (IPSN 2023)
Best Paper Runner-Up Award
Neural Weight Virtualization (MobiSys 2020)
Weight Separation (PerCom 2022)
Deep Functional Network (IPSN 2021)
Real-time Inference and Training of Machine Learning Algorithms
Intelligent Robot Systems
Smart Sensing Systems
SubFlow (RTAS 2020)
Multi-Device, Multi-Sensing, and Multi-Modal LearningÂ
Distributed Learning
Intermittent (Battery-less) Computing
On-demand Federated Learning (AISTATS 2024)
Intermittent Learning (IMWUT 2019/UbiComp 2020)
Best Presentation Award
Connected and Networked Intelligence
Learning in the wild (AIChallengeIoT 2020)
Best Paper Award
Generation and Optimization of Efficient Program Code and Executable Binary for Deep Learning Models
Bayesian Code Diffusion for Efficient Automatic Deep Learning Program Optimization (OSDI 2025)
AI and Machine Learning on Microcontrollers with Extreme Resource Constraints
NeuroZero (SenSys 2019)
Deep Controller Theory
Knowledge Transfer Between Embedded Controllers (DCOSS 2018)
Next-generation Neural Memory Architectures and Algorithms
Neural Memory for Machine Reasoning and Artificial General Intelligence
Enabling secure and confidential deep learning (machine learning) on resource-constrained embedded/mobile/IoT devices
We are very grateful for the support and assistance with our research and projects.
National Research Foundation
Institute for Information & communication Technology Planning & evaluation
Agency for Defense Development
Electronics and Telecommunications Research Institute
Korea Technology and Information Promotion Agency for SMEs
Korea Aerospace Research institute
Ulsan National Institute of Science and Technology
Hyundai Motor Company
Samsung Electronics
LG Electronics