Selected Presentations
You can also find the full list of my presentations on
CV.
October 15, 2022
Presentation, INFORMS 18th Workshop on Data Mining & Decision Analytics, Phoenix, Arizona, United States
Early detection of Alzheimer’s Disease (AD) is crucial to ensure timely interventions and optimize treatment outcomes for patients. While integrating multi-modal neuroimages, such as MRI and PET, has shown great promise, limited research has been done to effectively handle incomplete multi-modal image datasets in the integration. To this end, we propose a deep learning-based framework that employs Mutual Knowledge Distillation (MKD) to jointly model different sub-cohorts based on their respective available image modalities. In MKD, the model with more modalities (e.g., MRI and PET) is considered a teacher while the model with fewer modalities (e.g., only MRI) is considered a student. Our proposed MKD framework includes three key components: First, we design a teacher model that is student-oriented, namely the Student-oriented Multi-modal Teacher (SMT), through multi-modal information disentanglement. Second, we train the student model by not only minimizing its classification errors but also learning from the SMT teacher. Third, we update the teacher model by transfer learning from the student’s feature extractor because the student model is trained with more samples. Evaluations on Alzheimer’s Disease Neuroimaging Initiative (ADNI) datasets highlight the effectiveness of our method. Our work demonstrates the potential of using AI for addressing the challenges of incomplete multi-modal neuroimage datasets, opening new avenues for advancing early AD detection and treatment strategies.
October 15, 2022
Presentation, INFORMS 17th Workshop on Data Mining & Decision Analytics, Indianapolis, Indiana, United States
Early diagnosis of Alzheimer’s disease (AD) is an important task that facilitates the development of treatment and prevention strategies and may potentially improve patient outcomes. Neuroimaging has shown great promise, including the amyloid-PET which measures the accumulation of amyloid plaques in the brain – a hallmark of AD. It is desirable to train end-to-end deep learning models to predict the progression of AD for individuals at early stages based on 3D amyloid-PET. However, commonly used models are trained in a fully supervised learning manner and they are inevitably biased toward the given label information. To this end, we propose a self-supervised contrastive learning method to predict AD progression with 3D amyloid-PET. It uses unlabeled data to capture general representations underlying the images. As the downstream task is given as classification, unlike the general self-supervised learning problem that aims to generate task-agnostic representations, we also propose a loss function to utilize the label information in the pre-training. To demonstrate the performance of our method, we conducted experiments on the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset. The results confirmed that the proposed method is capable of providing appropriate data representations, resulting in accurate classification.
October 24, 2021
Presentation, 2021 INFORMS Annual Meeting,
Semi-supervised learning methods have been suffered from performance degradation when the class distributions of labeled and unlabeled data are diferrent. Even if previous studies have tackled the problem by removing the unncessary mismatch data. They might lose the basic information that all data share regardless of class. To this end, we propose to apply self-supervised learning to leverage the whole unlabeled data. We also propose a loss function to use in-distribution data as positive examples. We evaluate our method on image classification datasets under various mismatch ratios. The results show that our method produces good representation and improves classification accuracy.
June 02, 2021
Presentation, 2021 Korean Institute of Industrial Engineers, Jeju-do, South Korea
In the recent competitive landscape of the mobile phone market, companies are striving for dominance by incorporating advanced features such as 5G, foldable screens, and high-resolution cameras. This trend has increased product complexity and necessitated additional inspection processes to verify these enhanced functionalities, leading to the expansion of inspection equipment in factories. This continuous addition imposes a significant burden on companies. This study proposes a process to enhance productivity without the need for additional production lines and inspection equipment by estimating some inspection items traditionally dependent on measurement and inspection through modeling. Firstly, we utilize clustering models to group similar inspection items and select the optimal key items for each group. There is a high correlation among inspection items in the production process, and through this method, we identify only the essential inspection items that must be measured. Secondly, we train a model to predict the remaining inspection items using the key inspection items. Ultimately, the prediction values are used to forecast product defects. The effectiveness of the proposed method is demonstrated by collecting and analyzing real radio frequency inspection data.
October 20, 2019
Presentation, 2019 INFORMS Annual Meeting, Seattle, Washington, United States
Prediction of equipment failures is an important issue in various industries. Also, identifying the causes of the failures can be very helpful in determining how to deal with it. In this study, we propose an attention mechanism based neural network model that yields high performance to predict the failures with interpretability. The model provides attention distributions of sensor-level and segment-level denoting which sensor and segment contribute to the prediction. We evaluate the performance of the proposed method through the real multi-channel sensor data collected from the vehicle engine.
June 17, 2018
Presentation, 2018 INFORMS International Confererence, Taipei, Taiwan
Unexpected breakdown of the equipment significantly reduces productivity. The development of monitoring system that can detect abnormal conditions early is essential. In this study, we propose the multivariate monitoring method based on a convolutional autoencoder algorithm that can effectively reconstruct sensor data. The proposed monitoring method identifies the equipment status and detects critical variables that cause the alarms. We evaluate the performance of the proposed method with actual sensor data collected from construction equipment.
November 19, 2016
Presentation, Industrial Engineering & Management Science Conference, Seoul, Korea
Parts demand forecasting is a crucial technique for determining production quantities, pricing strategies, and inventory management. Unlike the demand patterns for construction equipment and automobiles, the demand for spare parts exhibits an intermittent pattern. The intervals between demand occurrences are irregular, and the quantity of demand varies significantly. These characteristics pose significant challenges for traditional time series models in capturing the demand patterns for parts. In this study, we propose a forecasting model designed to address the time series problems characterized by intermittent patterns. We demonstrate the effectiveness of the proposed model by applying it to real parts sales data for demand forecasting.
