A5101622080- Reservoir Engineering and Simulation Methods
- Hydraulic Fracturing and Reservoir Analysis
- Explainable Artificial Intelligence (XAI)
- Drilling and Well Engineering
- Domain Adaptation and Few-Shot Learning
- Advanced Neural Network Applications
- Breast Cancer Treatment Studies
- AI in cancer detection
- Adversarial Robustness in Machine Learning
- Radiomics and Machine Learning in Medical Imaging
- Brain Tumor Detection and Classification
- Seismic Imaging and Inversion Techniques
- Topic Modeling
- Mineral Processing and Grinding
- Nonmelanoma Skin Cancer Studies
- Handwritten Text Recognition Techniques
- Multimodal Machine Learning Applications
- Water resources management and optimization
- Skin Protection and Aging
- Marine and fisheries research
- Cholangiocarcinoma and Gallbladder Cancer Studies
- Advanced Steganography and Watermarking Techniques
- Electrolyte and hormonal disorders
- QR Code Applications and Technologies
- Visual Attention and Saliency Detection
Seoul National University
2016-2021
Samsung SDS (South Korea)
2021
Samsung (United States)
2021
Samsung (South Korea)
2021
Catholic University of Korea
1999-2011
Research & Development Institute
2010
Summary Decline–curve analysis (DCA) is an easy and fast empirical regression method for predicting future well production. However, applying DCA to shale–gas wells limited by long transient flow, a unique completion design, high–density drilling. Recently, short-term-memory (LSTM) algorithm has been widely applied the prediction of time–series data. Because shale–gas–production data are data, LSTM can be predict After information 332 in Alberta, Canada, obtained from commercial database,...
Abstract The achievement of the pathologic complete response (pCR) has been considered a metric for success neoadjuvant chemotherapy (NAC) and powerful surrogate indicator risk recurrence long-term survival. This study aimed to develop multimodal deep learning model that combined clinical information pretreatment MR images predicting pCR NAC in patients with breast cancer. retrospective cohort consisted 536 invasive cancer who underwent pre-operative NAC. We developed fuse high-dimensional...
Increasing demands for understanding the internal behavior of convolutional neural networks (CNNs) have led to remarkable improvements in explanation methods. Particularly, several class activation mapping (CAM) based methods, which generate visual maps by a linear combination from CNNs, been proposed. However, majority methods lack clear theoretical basis on how they assign coefficients combination. In this paper, we revisit intrinsic linearity CAM with respect maps; construct an model CNN...
Ensemble Kalman filter (EnKF) has been widely studied due to its excellent recursive data processing, dependable uncertainty quantification, and real-time update. However, many previous works have shown poor characterization results on channel reservoirs with non-Gaussian permeability distribution, which do not satisfy the Gaussian assumption of EnKF algorithm. To meet assumption, normal score transformation can be applied ensemble parameters. Even though this preserves initial distribution...
Ensemble-based analyses are useful to compare equiprobable scenarios of the reservoir models. However, they require a large suite models cover high uncertainty in heterogeneous and complex For stable convergence ensemble Kalman filter (EnKF), increasing size can be one solutions, but it causes computational cost large-scale systems. In this paper, we propose preprocessing good initial model selection reduce size, then, EnKF is utilized predict production performances stochastically. scheme,...
Open-vocabulary object detection (OVOD) aims to recognize novel objects whose categories are not included in the training set. In order classify these unseen classes during training, many OVOD frameworks leverage zero-shot capability of largely pretrained vision and language models, such as CLIP. To further improve generalization on classes, several approaches proposed additionally train with pseudo region labeling external data sources that contain a substantial number category labels...
Reservoir characterization is a process of making models, which reliably predict reservoir behaviors. Ensemble Kalman filter (EnKF) one the fine methods for with many advantages. However, it hard to get trustworthy results in discrete grid system ensuring preservation channel properties. There have been schemes such as cosine transform (DCT) and facies ratio (PFR) improvement reservoirs characterization. These are mostly applied 2D cases, but cannot present satisfactory 3D gas an aquifer...
Ensemble Kalman filter (EnKF) uses recursive updates for data assimilation and provides dependable uncertainty quantification. However, it requires high computing cost. On the contrary, ensemble smoother (ES) assimilates all available simultaneously. It is simple fast, but prone to showing two key limitations: overshooting divergence. Since channel fields have non-Gaussian distributions, challenging characterize them with conventional based history matching methods. In many cases, a large...
Numerous interpretability methods have been developed to visually explain the behavior of complex machine learning models by estimating parts input image that are critical for model's prediction. We propose a general pipeline enhancing visual explanations using transformations (EVET). EVET considers original refine region based on an intuitive rationale estimated be important in variously transformed inputs is more important. Our proposed applicable existing explanation without modification....
s and Programme: EUROANAESTHESIA 2011: The European Anaesthesiology Congress: Transfusion Haemostasis
This paper presents an innovative data-integration that uses iterative-learning method, a deep neural network (DNN) coupled with stacked autoencoder (SAE) to solve issues encountered many-objective history matching. The proposed method consists of DNN-based inverse model SAE-encoded static data and iterative updates supervised-learning are based on distance-based clustering schemes. DNN functions as results in encoded flattened data, while SAE, pre-trained network, successfully reduces...
Reservoir characterization is a process to make dependable reservoir models using available information. There are promising ensemble-based methods such as ensemble Kalman filter (EnKF), smoother (ES), and with multiple data assimilation (ES-MDA). ES-MDA an iterative version of ES inflated covariance matrix measurement errors. It provides efficient consistent global updates compared EnKF ES. Ensemble-based method might not work properly for channel reservoirs because its parameters highly...
Ensemble Kalman filter (EnKF) is one of the powerful optimization schemes for production data history matching in petroleum engineering. It provides promising characterization results and dependable future prediction performances. However, it needs high computational cost due to its recursive updating procedures. smoother (ES), which updates all available observation at once, has calculation efficiency but tends give unreliable compared with EnKF. Particularly, challenging channel...
The recent segmentation foundation model, Segment Anything Model (SAM), exhibits strong zero-shot capabilities, but it falls short in generating fine-grained precise masks. To address this limitation, we propose a novel image matting called ZIM, with two key contributions: First, develop label converter that transforms labels into detailed matte labels, constructing the new SA1B-Matte dataset without costly manual annotations. Training SAM enables to generate masks while maintaining its...
Abstract Background The aim of this study was to develop a machine learning(ML) based model accurately predict pathologic complete response(pCR) neoadjuvant chemotherapy(NAC) using pretreatment clinical and pathological characteristics electronic medical record(EMR) data in breast cancer(BC). Methods EMR from patients diagnosed with early locally advanced BC who received NAC followed by curative surgery were reviewed. A total 16 selected ML model. We practiced six models default settings for...
Increasing demands for understanding the internal behavior of convolutional neural networks (CNNs) have led to remarkable improvements in explanation methods. Particularly, several class activation mapping (CAM) based methods, which generate visual maps by a linear combination from CNNs, been proposed. However, majority methods lack clear theoretical basis on how they assign coefficients combination. In this paper, we revisit intrinsic linearity CAM with respect maps; construct an model CNN...