- Reinforcement Learning in Robotics
- Anomaly Detection Techniques and Applications
- Domain Adaptation and Few-Shot Learning
- Advanced Control Systems Optimization
- Fault Detection and Control Systems
- Brain Tumor Detection and Classification
- Mineral Processing and Grinding
- Oral and Maxillofacial Pathology
- Multimodal Machine Learning Applications
- Industrial Vision Systems and Defect Detection
- Face and Expression Recognition
- Dementia and Cognitive Impairment Research
- Dental Radiography and Imaging
- Network Security and Intrusion Detection
- Non-Destructive Testing Techniques
- Advanced Chemical Sensor Technologies
- Artificial Immune Systems Applications
- Neuroinflammation and Neurodegeneration Mechanisms
- Advanced Statistical Process Monitoring
- Time Series Analysis and Forecasting
- Gait Recognition and Analysis
- Evolutionary Algorithms and Applications
- Advanced Neural Network Applications
- Radiomics and Machine Learning in Medical Imaging
- Smart Parking Systems Research
Georgia Institute of Technology
2022-2025
Korea University
2019-2021
In pixel-based deep reinforcement learning (DRL), representations of states that change because an agent's action or interaction with the environment poses a critical challenge in improving data efficiency. Recent data-efficient DRL studies have integrated self-supervised (SSL) and augmentation to learn state from given interactions. However, some methods difficulties explicitly capturing evolving selecting augmentations for appropriate reward signals. Our goal is inherent dynamics...
Detecting an anomaly in multichannel signal data is a challenging task various domains. It should take into account the cross-channel relationship and temporal within each channel. Moreover, high-dimensional making it difficult to gather sufficient abnormal labels. Consequently, unsupervised reconstruction-based detection methods have been applied successfully many studies. However, they lose valuable channel information inherent reconstruction errors by merely averaging for both time, then...
Early diagnosis of Alzheimer's disease (AD) is an important task that facilitates the development treatment and prevention strategies may potentially improve patient outcomes. Neuroimaging has shown great promise, including amyloid-PET which measures accumulation amyloid plaques in brain - a hallmark AD. It desirable to train end-to-end deep learning models predict progression AD for individuals at early stages based on 3D amyloid-PET. However, commonly used are trained fully supervised...
Defect patterns exhibited in wafer bin maps (WBMs) can provide essential clues about critical process failures to field engineers. In modern manufacturing processes, the automatic WBM defect pattern classification is for yield improvement. Although it difficult collect sufficient labels while a lot of unlabeled data given, most existing studies have mainly used only labeled data. Moreover, out-of-distribution (OOD) WBMs are inevitably collected. It degrades performance semi-supervised...
Early diagnosis of Alzheimer’s disease (AD) is an important task that facilitates the development treatment and prevention strategies, may potentially improve patient outcomes. Neuroimaging has shown great promise, including amyloid-PET, which measures accumulation amyloid plaques in brain—a hallmark AD. It desirable to train end-to-end deep learning models predict progression AD for individuals at early stages based on 3D amyloid-PET. However, commonly used are trained a fully supervised...
Semi-supervised learning methods have shown promising results in solving many practical problems when only a few labels are available. The existing assume that the class distributions of labeled and unlabeled data equal; however, their performances significantly degraded distribution mismatch scenarios where out-of-distribution (OOD) exist data. Previous safe semi-supervised studies addressed this problem by making OOD less likely to affect training based on However, even if effectively...
Early detection of Alzheimer's Disease (AD) is crucial for timely interventions and optimizing treatment outcomes. Despite the promise integrating multimodal neuroimages such as MRI PET, handling datasets with incomplete modalities remains under-researched. This phenomenon, however, common in real-world scenarios not every patient has all due to practical constraints cost, access, safety concerns. We propose a deep learning framework employing cross-modal Mutual Knowledge Distillation (MKD)...