A5005505577- Topic Modeling
- Natural Language Processing Techniques
- Advanced Text Analysis Techniques
- Advanced Malware Detection Techniques
- Computational Drug Discovery Methods
- Speech and dialogue systems
- Network Security and Intrusion Detection
- Sentiment Analysis and Opinion Mining
- Robotics and Automated Systems
- Web Data Mining and Analysis
- Crystallography and molecular interactions
- Text and Document Classification Technologies
- Anomaly Detection Techniques and Applications
- Crystallization and Solubility Studies
- Hemodynamic Monitoring and Therapy
- Neural Networks and Applications
- Cardiac, Anesthesia and Surgical Outcomes
- Computational and Text Analysis Methods
- Machine Learning in Materials Science
- Context-Aware Activity Recognition Systems
- Image Retrieval and Classification Techniques
- Speech Recognition and Synthesis
- Spam and Phishing Detection
- Advanced Image and Video Retrieval Techniques
- Machine Learning in Bioinformatics
Chungbuk National University
2021-2025
Soonchunhyang University
2016-2021
Korea Advanced Institute of Science and Technology
2011-2018
Daejeon University
2016
Naver (South Korea)
2016
As the number of textual data is exponentially increasing, it becomes more important to develop models analyze text automatically. The texts may contain various labels such as gender, age, country, sentiment, and so forth. Using bring benefits some industrial fields, many studies classification have appeared. Recently, Convolutional Neural Network (CNN) has been adopted for task shown quite successful results. In this paper, we propose convolutional neural networks sentiment classification....
Hypotensive events in the initial stage of anesthesia can cause serious complications patients after surgery, which could be fatal. In this study, we intended to predict hypotension tracheal intubation using machine learning and deep techniques one minute advance. Meta models, such as random forest, extreme gradient boosting (Xgboost), especially convolutional neural network (CNN) model (DNN), were trained occurring between incision, data from four minutes before intubation. Vital records...
Cocrystals are of much interest in industrial application as well academic research, and screening suitable coformers for active pharmaceutical ingredients is the most crucial challenging step cocrystal development. Recently, machine learning techniques attracting researchers many fields including research such quantitative structure-activity/property relationship. In this paper, we develop models to predict formation. We extract descriptor values from simplified molecular-input line-entry...
Predicting the impact of mutations on protein-ligand binding affinity is crucial in drug discovery, particularly addressing resistance and repurposing existing drugs. Current structure-based methods, including deep learning physics-based computational techniques, are constrained by their dependence known complex structures. The lack structural data for mutated proteins, coupled with potential to alter protein conformational states, poses challenges reliable predictions. To address challenge,...
The prevalence of non-executable malware is on the rise, presenting a major threat to users, including public institutions and corporations. While extensive research has been conducted detecting threats, there noticeable gap in studying document-type compared with executable files. proposed model will solve this by classifying families using script codes, tags, write documents languages execute malicious functions. These codes offer insights into how was constructed operates victim’s system....
Abstract Pre-trained language models have brought significant performance improvements in many natural understanding tasks. Domain-adaptive models, which are trained with a specific domain corpus, exhibit high their target domains. However, pre-training these large amount of domain-specific data requires substantial computational budget and resources, necessitating the development efficient methods. In this paper, we propose novel subset selection method called AlignSet, extracts an...
With increasing amount of data, the threat malware keeps growing recently. The malicious actions embedded in nonexecutable documents especially (e.g., PDF files) can be more dangerous, because it is difficult to detect and most users are not aware such type attacks. In this paper, we design a convolutional neural network tackle detection on files. We collect benign files manually label byte sequences within intensively examine structure input data illustrate how proposed based...
Scene text detection is the task of detecting word boxes in given images. The accuracy has been greatly elevated using deep learning models, especially convolutional neural networks. Previous studies commonly aimed at developing more accurate but their models became computationally heavy and worse efficiency. In this paper, we propose a new efficient model for detection. proposed model, namely Compact Accurate Text detector (CAST), consists MobileNetV2 as backbone balanced decoder. Unlike...
Malaria remains by far one of the most threatening and dangerous illnesses caused plasmodium falciparum parasite. Chloroquine (CQ) first-line artemisinin-based combination treatment (ACT) have long been drug choice for controlling malaria; however, emergence CQ-resistant artemisinin resistance parasites is now present in areas where malaria endemic. In this work, we developed five machine learning models to predict antimalarial bioactivities a against from features (i.e., molecular...
Gati Martin, Medard Edmund Mswahili, Young-Seob Jeong, Jeong Young-Seob. Proceedings of the 2022 Conference North American Chapter Association for Computational Linguistics: Human Language Technologies. 2022.
Data-driven approaches employ the stochastic process that is simulated using random numbers. The number can be thought of as an unpredictable value without bias or correlation, so it essentially impossible to design a system generates `real' There have been studies, therefore, aimed at developing pseudo generator, where not perfectly random, but practically useful. In this paper, we propose new for generation. recurrent neural networks with long short-term memery (LSTM) units are used mimic...
Anesthesia induction is associated with frequent blood pressure fluctuation such as hypotension and hypertension. If it possible to precisely predict a few minutes ahead, anesthesiologists can proactively give anesthetic management before patients develop hemodynamic problem. The objective of this study real-time model for predicting 3-min-ahead from the start anesthesia surgical incision. We used only vital signs anesthesia-related data obtained during anesthesia-induction phase designed...
BACKGROUND Tuberculous meningitis (TBM) is the most severe form of tuberculosis, but differentiating between diagnosis TBM and viral (VM) difficult. Thus, we have developed machine-learning modules for from VM. MATERIAL AND METHODS For training data, confirmed or probable VM cases were retrospectively collected five teaching hospitals in Korea January 2000 - July 2018. Various algorithms used training. The tested by leave-one-out cross-validation. Four residents two infectious disease...
The evolution of the Internet has increased amount information that is expressed by people on different platforms. This can be product reviews, discussions forums, or social media Accessibility these opinions and peoples feelings open door to opinion mining sentiment analysis. As language speech technologies become more advanced, many languages have been used best models obtained. However, due linguistic diversity lack datasets, African left behind. In this study, using current...
Pharmaceutical cocrystals of pelubiprofen (PF) were discovered for the first time. 16 candidates to form with PF selected via ANN model and p K a rule.
The growing demand for coffee has led to the development industry. Since defect beans affect taste of coffee, it is essential select them improve quality coffee. This basically a classification task that predicts appropriate types in given beans. When person working manually this classification, can be affected by human condition and disadvantage taking long time. There have been few studies utilized data-driven method predict analyzing images, they commonly used convolutional neural network...
Malaria continues to pose a significant global health burden despite concerted efforts combat it. In 2020, nearly half of the world’s population faced risk malaria, underscoring urgency innovative strategies tackle this pervasive threat. One major challenges lies in emergence resistance parasites existing antimalarial drugs. This challenge necessitates discovery new, effective treatments capable combating Plasmodium parasite at various stages its life cycle. Advanced computational approaches...
End stage renal disease (ESRD) is the last of chronic kidney that requires dialysis or a transplant to survive. Many studies reported higher risk mortality in ESRD patients compared with without ESRD. In this paper, we develop model predict postoperative complications, major cardiac event, for who underwent any type surgery. We compare several widely-used machine learning models through experiments our collected data yellow size 3220, and achieved F1 score 0.797 random forest model. Based on...
This research is aimed at predicting the physical stability for amorphous solid dispersion by utilizing deep learning methods. We propose a prediction model that effectively learns from small dataset imbalanced in terms of class. In order to overcome imbalance problem, our performs hybrid sampling which combines synthetic minority oversampling technique (SMOTE) algorithm with edited nearest neighbor (ENN) and reduces dimensionality using principal component analysis (PCA) during data...
Intelligent attacks using document-based malware that exploit vulnerabilities in document viewing software programs or file structure are increasing rapidly. There many cases of PDF (portable format) proportion to its usage. We provide in-depth analysis on and JavaScript content embedded PDFs. Then, we develop the diverse feature set encompassing metadata such as size, version, encoding method keywords, features object names, readable strings JavaScript. When diverse, it is hard adversarial...