Minhyeok Lee

Yoojoong Kim

Donghyun Tae

Wonsuk Kim

Sunho Choi

Insoo Kim

Jangwon Seo

Jaeyoon Kim

Sunyoung Lee

Jiyoon Bae

Ghudae Sim

Hyung Bin Yun

Dong Hyun Lim

Soojeong Kim

Jeongeun Kim

Eunseong Seo

Taeho Lee

Wook Lee

Seoyoung Sim

Seunghwan Lee

Noori Yun

Seungyun Baek

Chanjong Yu

Hyeju Son

Yookyung Koh

The Complex Data Analytics Lab focuses on developing and applying computational methods for the analysis of complex, large-scale data. It addresses challenges posed by the explosion of big data, including finance, sensor data from IoT systems, and genome information. The lab develops advanced automatic systems using AI, machine learning, and current computational technology to analyze big data and enhance the quality of life.

복잡 데이터 분석 연구실은 복잡하고 대규모의 데이터 분석을 위한 전산 방법 개발 및 적용에 중점을 둔다. 금융, IoT 시스템의 센서 데이터, 유전체 정보를 포함한 빅데이터 문제를 해결하고, 인공지능, 머신러닝, 최신 컴퓨팅 기술을 기반으로 혁신적인 분석 도구를 개발하여 삶의 질을 향상시키는 고급 자동 시스템 개발을 목표로 한다.

Deriving Optimal Atomic Layer Deposition Process Conditions Using Machine Learning

Adversarial environment design for crowd navigation based on deep reinforcement learning

Automated BigSMILES conversion workflow and dataset for homopolymeric macromolecules

Prioritizing Disease Diagnosis in Neonatal Cohorts through Multivariate Survival Analysis: A Nonparametric Bayesian Approach

A Multi-Stage Deep Learning Approach Incorporating Text-Image and Image-Image Comparisons for Cheapfake Detection

Stabilized GAN Models Training with Kernel-Histogram Transformation and Probability Mass Function Distance

ctGAN: combined transformation of gene expression and survival data with generative adversarial network

Generative Adversarial Soft Actor–Critic

CTIVA: Censored time interval variable analysis

Clinical significance of urinary exosomal microRNAs in patients with IgA nephropathy

Deep learning model with L1 penalty for predicting breast cancer metastasis using gene expression data

Corrigendum: Development and validation of urinary exosomal microRNA biomarkers for the diagnosis of acute rejection in kidney transplant recipients

ICEGAN: Inverse Covariance Estimating Generative Adversarial Network

Grid-wise simulation acceleration of the electromagnetic fields of 2D optical devices using super-resolution

Score-guided generative adversarial networks

Conversion of Automated 12-Lead Electrocardiogram Interpretations to OMOP CDM Vocabulary

Inverse design of nanophotonic devices using generative adversarial networks

Deep reinforcement learning with a critic-value-based branch tree for the inverse design of two-dimensional optical devices

Searching similar weather maps using convolutional autoencoder and satellite images

Simulator acceleration and inverse design of fin field-effect transistors using machine learning

Non-Invasive Diagnosis for Acute Rejection Using Urinary mRNA Signature Reflecting Allograft Status in Kidney Transplantation

Urinary mRNA Signatures as Predictors of Renal Function Decline in Patients With Biopsy-Proven Diabetic Kidney Disease

Estimation with Uncertainty via Conditional Generative Adversarial Networks

Validation of deep learning natural language processing algorithm for keyword extraction from pathology reports in electronic health records

Network estimation for censored time-to-event data for multiple events based on multivariate survival analysis

Simulation Acceleration for Transmittance of Electromagnetic Waves in 2D Slit Arrays Using Deep Learning

Phenotype and Molecular Signature of CD8+ T Cell Subsets in T cell-Mediated Rejections After Kidney Transplantation

Improved Recurrent Generative Adversarial Networks with Regularization Techniques and a Controllable Framework

Clinical Implication of Concordant or Discordant Genomic Profiling Between Primary and Matched Metastatic Tissues in Patients With Colorectal Cancer

Serum Biomarkers From Cell-Based Assays for AhRL and MIS Strongly Predicted the Future Development of Diabetes in a Large Community-Based Prospective Study in Korea

Portfolio Management via Two-stage Deep Learning with A Joint Cost

Prediction of survival risks with adjusted gene expression through risk-gene networks

Controllable Generative Adversarial Network

Regression-Based Network Estimation for High-Dimensional Genetic Data

mEBT: multiple-matching Evidence-based Translator of Murine Genomic Responses for Human Immunity Studies.

GAIT: Gene expression Analysis for Interval Time

The estimation of probability distribution for factor variables with many categorical values

A Comparison of Two-Stage Approaches Based on Penalized Regression for Estimating Gene Networks

Estimation of directed subnetworks in ultra high dimensional data for gene network problems

RASA: Robust Alternative Splicing Analysis for Human Transcriptome Arrays

Mutual Information between Discrete Variables with Many Categories using Recursive Adaptive Partitioning.

A hybrid approach of gene sets and single genes for the prediction of survival risks with gene expression data.

Evidence-based translation for the genomic responses of murine models for the study of human immunity.

Mice are not men.

Density estimation on multivariate censored data with optional Pólya tree

Gene response in mouse models poorly mimics severe human inflammatory diseases

Development of a genomic metric that can be rapidly used to predict clinical outcome in severely injured trauma patients

Coding SNPs as intrinsic markers for sample tracking in large-scale transcriptome studies

JETTA: junction and exon toolkits for transcriptome analysis

Knowledge-based reconstruction of mRNA transcripts with short sequencing reads for transcriptome research

A genomic storm in critically injured humans

Distinctive responsiveness to stromal signaling accompanies histologic grade programming of cancer cells

Human transcriptome array for high-throughput clinical studies

Knowledge-based analysis of microarrays for the discovery of transcriptional regulation relationships

A dynamic network of transcription in LPS-treated human subjects

Using high-density exon arrays to profile gene expression in closely related species

Bisphenol A induces a profile of tumor aggressiveness in high-risk cells from breast cancer patients

Genomic responses in mouse models poorly mimic human inflammatory diseases

Inflammation and Host Response to Injury, Large Scale Collaborative Research Program. Genomic responses in mouse models poorly mimic human inflammatory diseases

Mice are not men

Blood transcriptomic biomarkers in adult primary care patients with major depressive disorder undergoing cognitive behavioral therapy

Controllable generative adversarial network

A survey of missing data imputation using generative adversarial networks

Portfolio management via two-stage deep learning with a joint cost

Indoor semantic segmentation for robot navigating on mobile

Regularization methods for generative adversarial networks: An overview of recent studies

Stock prices prediction using the title of newspaper articles with Korean natural language processing

Prediction of information propagation in a drone network by using machine learning

Inflammation and host response to injury large-scale collaborative research program. Human transcriptome array for high-throughput clinical studies

wenzhong xiao

Ronald Tompkins

Junhee Seok

A cornerstone of modern biomedical research is the use of mouse models to explore basic pathophysiological mechanisms, evaluate new therapeutic approaches, and make go or no-go decisions to carry new drug candidates forward into clinical trials. Systematic studies evaluating how well murine models mimic human inflammatory diseases are nonexistent. Here, we show that, although acute inflammatory stresses from different etiologies result in highly similar genomic responses in humans, the responses in corresponding mouse models correlate poorly with the human conditions and also, one another. Among genes changed significantly in humans, the murine orthologs are close to random in matching their human counterparts (e.g., R 2 between 0.0 and 0.1). In addition to improvements in the current animal model systems, our study supports higher priority for translational medical research to focus on the …

Human survival from injury requires an appropriate inflammatory and immune response. We describe the circulating leukocyte transcriptome after severe trauma and burn injury, as well as in healthy subjects receiving low-dose bacterial endotoxin, and show that these severe stresses produce a global reprioritization affecting >80% of the cellular functions and pathways, a truly unexpected “genomic storm.” In severe blunt trauma, the early leukocyte genomic response is consistent with simultaneously increased expression of genes involved in the systemic inflammatory, innate immune, and compensatory antiinflammatory responses, as well as in the suppression of genes involved in adaptive immunity. Furthermore, complications like nosocomial infections and organ failure are not associated with any genomic evidence of a second hit and differ only in the magnitude and duration of this genomic reprioritization. The …

A vibrant discussion of the merits and limitations of animal models is long overdue. The limitation of space precludes addressing many of the questionable approaches and statements by Takao and Miyakawa (1). Despite the different approaches used by Takao and Miyakawa (1), their results actually support the conclusion that “Genomic responses in mouse models poorly mimic human inflammatory diseases”(2). This can be best understood using the example provided by the authors:“13,586 and 3,116 genes are changed (P< 0.05 and fold-change> 1.2) in human burn conditions and mouse models of infection, respectively, and 1,992 among them are commonly changed in both humans and mice, among which 1,608 genes changed to the same direction (Fig. 2E)”(1). Therefore, among the 13,586 genes changed in the human disease, the model can reflect only 1,608 (or 12%) correctly to some extent. The …

Breast cancer outcome is highly variable. Whether inadvertent exposure to environmental xenobiotics evokes a biological response promoting cancer aggressiveness and a higher probability of tumor recurrence remains unknown. To determine specific molecular alterations which arise in high-risk breast tissue in the presence of the ubiquitous xenoestrogen, bisphenol A (BPA), we used nonmalignant random periareolar fine-needle aspirates in a novel functional assay. Early events induced by BPA in epithelial-stromal cocultures derived from the contralateral tissue of patients with breast cancer included gene expression patterns which facilitate apoptosis evasion, endurance of microenvironmental stress, and cell cycle deregulation without a detectable increase in cell numbers. This BPA response profile was significantly associated with breast tumors characterized by high histologic grade ( P < 0.001) and …

A 6.9 million-feature oligonucleotide array of the human transcriptome [Glue Grant human transcriptome (GG-H array)] has been developed for high-throughput and cost-effective analyses in clinical studies. This array allows comprehensive examination of gene expression and genome-wide identification of alternative splicing as well as detection of coding SNPs and noncoding transcripts. The performance of the array was examined and compared with mRNA sequencing (RNA-Seq) results over multiple independent replicates of liver and muscle samples. Compared with RNA-Seq of 46 million uniquely mappable reads per replicate, the GG-H array is highly reproducible in estimating gene and exon abundance. Although both platforms detect similar expression changes at the gene level, the GG-H array is more sensitive at the exon level. Deeper sequencing is required to adequately cover low-abundance …

An objective, laboratory-based diagnostic tool could increase the diagnostic accuracy of major depressive disorders (MDDs), identify factors that characterize patients and promote individualized therapy. The goal of this study was to assess a blood-based biomarker panel, which showed promise in adolescents with MDD, in adult primary care patients with MDD and age-, gender-and race-matched nondepressed (ND) controls. Patients with MDD received cognitive behavioral therapy (CBT) and clinical assessment using self-reported depression with the Patient Health Questionnaire–9 (PHQ-9). The measures, including blood RNA collection, were obtained before and after 18 weeks of CBT. Blood transcript levels of nine markers of ADCY3, DGKA, FAM46A, IGSF4A/CADM1, KIAA1539, MARCKS, PSME1, RAPH1 and TLR7, differed significantly between participants with MDD (N= 32) and ND controls (N= 32) at …

Recently introduced generative adversarial networks (GANs) have been shown numerous promising results to generate realistic samples. In the last couple of years, it has been studied to control features in synthetic samples generated by the GAN. Auxiliary classifier GAN (ACGAN), a conventional method to generate conditional samples, employs a classification layer in discriminator to solve the problem. However, in this paper, we demonstrate that the auxiliary classifier can hardly provide good guidance for training of the generator, where the classifier suffers from overfitting. Since the generator learns from classification loss, such a problem has a chance to hinder the training. To overcome this limitation, here, we propose a controllable GAN (ControlGAN) structure. By separating a feature classifier from the discriminator, the classifier can be trained with data augmentation technique, which can support to make a …

Objective: Many patients have complicated recoveries following severe trauma due to the development of organ injury. Physiological and anatomical prognosticators have had limited success in predicting clinical trajectories. We report on the development and retrospective validation of a simple genomic composite score that can be rapidly used to predict clinical outcomes. Design: Retrospective cohort study. Setting: Multi-institutional level 1 trauma centers. Patients: Data were collected from 167 severely traumatized (injury severity score> 15) adult (18–55 yr) patients. Methods: Microarray-derived genomic data obtained from 167 severely traumatized patients over 28 days were assessed for differences in messenger RNA abundance among individuals with different clinical trajectories. Once a set of genes was identified based on differences in expression over the entire study period, messenger RNA abundance from …

Recently, many deep learning models for missing data imputation have been studied. One of the most popular models is Generative Adversarial Networks (GANs), which generate plausible fake data through adversarial training. In this paper, we take a look at the architecture, objective of a generator and a discriminator, training method and loss function. After that, we can see what improvements have been made to each model. Moreover, we can easily compare several GAN-based models for missing data imputation.

Portfolio management is a series of processes that maximize returns and minimize risk by allocating assets efficiently. Along with the developments in machine learning technology, it has been studied to apply machine learning methods to prediction-based portfolio management. However, such methods have a few limitations. First, they do not consider the relations between assets for the prediction. In addition, the studies commonly focus on the prediction accuracy, neglecting the construction of portfolios. Furthermore, the methods have usually been evaluated with index data, which hardly represent actual prices to buy or sell an asset. To overcome these problems, Exchange Traded Funds (ETFs) are employed for base assets for the evaluation, and we propose a two-stage deep learning framework, called Grouped-ETFs Model (GEM), with a joint cost function. The GEM is designed to learn the features of inter …

Pathology reports contain the essential data for both clinical and research purposes. However, the extraction of meaningful, qualitative data from the original document is difficult due to the narrative and complex nature of such reports. Keyword extraction for pathology reports is necessary to summarize the informative text and reduce intensive time consumption. In this study, we employed a deep learning model for the natural language process to extract keywords from pathology reports and presented the supervised keyword extraction algorithm. We considered three types of pathological keywords, namely specimen, procedure, and pathology types. We compared the performance of the present algorithm with the conventional keyword extraction methods on the 3115 pathology reports that were manually labeled by professional pathologists. Additionally, we applied the present algorithm to 36,014 unlabeled …

In recent years, there have been many successes of using Deep Convolutional Neural Networks (DCNNs) in the task of pixel-level classification (also called “semantic image segmentation”). The advances in DCNN have led to the development of autonomous vehicles that can drive with no driver controls by using sensors like camera, LiDAR, etc. In this paper, we propose a practical method to implement autonomous indoor navigation based on semantic image segmentation using state-of-the-art performance model on mobile devices, especially Android devices. We apply a system called `Mobile DeepLabv3', which uses atrous convolution when applying semantic image segmentation by using MobileNetV2 as a network backbone. The ADE20K dataset is used to train our models specific to indoor environments. Since this model is for robot navigating, we re-label 150 classes into 20 classes in order to easily …

Background Understanding the transcriptional regulatory networks that map out the coordinated dynamic responses of signaling proteins, transcription factors and target genes over time would represent a significant advance in the application of genome wide expression analysis. The primary challenge is monitoring transcription factor activities over time, which is not yet available at the large scale. Instead, there have been several developments to estimate activities computationally. For example, Network Component Analysis (NCA) is an approach that can predict transcription factor activities over time as well as the relative regulatory influence of factors on each target gene. Results In this study, we analyzed a gene expression data set in blood leukocytes from human subjects administered with lipopolysaccharide (LPS), a prototypical inflammatory challenge, in the context of a reconstructed regulatory network …

Despite its short history, Generative Adversarial Network (GAN) has been extensively studied and used for various tasks, including its original purpose, i.e., synthetic sample generation. However, applying GAN to different data types with diverse neural network architectures has been hindered by its limitation in training, where the model easily diverges. Such a notorious training of GANs is well known and has been addressed in numerous studies. Consequently, in order to make the training of GAN stable, numerous regularization methods have been proposed in recent years. This paper reviews the regularization methods that have been recently introduced, most of which have been published in the last three years. Specifically, we focus on general methods that can be commonly used regardless of neural network architectures. To explore the latest research trends in the regularization for GANs, the methods are classified into several groups by their operation principles, and the differences between the methods are analyzed. Furthermore, to provide practical knowledge of using these methods, we investigate popular methods that have been frequently employed in state-of-the-art GANs. In addition, we discuss the limitations in existing methods and propose future research directions.

The efficient design of structures that exhibit desired properties is challenging across various engineering and scientific applications. Traditional methods employ experts in a specific domain to design new structures with desired properties. Then, simulations are performed for the designed structures to evaluate whether they show desired properties, and such a process is with until the structures exhibit desired properties. Advances in computing power and machine learning have made these simulations and optimizations faster, but challenges remain that the researchers must perform optimizations in each iteration, which generally takes time and cost. A new framework called inverse design has been studied to address the limitations. In inverse design, structures with desired properties can directly be constructed. In this work, as an inverse design framework, we introduce a controllable generative adversarial …

Non-quantitative data have a significant impact on the financial market as well as quantitative data. In this paper, we propose CNN model of stock price prediction using Korean natural language processing. In the case of Korean natural language processing research was not actively performed compared to English language. We converted Korean sentences into nouns and vectorized them using skip-grams to extract the characteristics of the words. Then, the vectorized word sentence was used as input data of the CNN model to predict the stock price after 5 days of trading day. Most models have more than 50% prediction accuracy for stock price up and down. The highest accuracy of the model was about 53%. Since the result is not considerable but meaningful, it shows the possibility of developing the stock price prediction model through Korean natural language processing in the future.

Drones cooperate with each other by transmitting and receiving packets. Therefore, it is important to conjecture the packet transmission rates within the network. However, the conventional methods are not suitable to describe the transmission patterns with satisfactory computing speed and accuracy. In this paper, we demonstrated that machine learning can successfully predict the transmission patterns in drone network. The packet transmission rates of a communication network with twenty drones were simulated, of which results were used to train the linear regression and Support Vector Machine with Quadratic Kernel (SVM-QK). We found out SVM-QK can precisely predict the communication between drones.

The simulation and design of electronic devices such as transistors is vital for the semiconductor industry. Conventionally, a device is intuitively designed and simulated using model equations, which is a time-consuming and expensive process. However, recent machine learning approaches provide an unprecedented opportunity to improve these tasks by training the underlying relationships between the device design and the specifications derived from the extensively accumulated simulation data. This study implements various machine learning approaches for the simulation acceleration and inverse-design problems of fin field-effect transistors. In comparison to traditional simulators, the proposed neural network model demonstrated almost equivalent results (R 2 = 0.99) and was more than 122,000 times faster in simulation. Moreover, the proposed inverse-design model successfully generated design …

CDAL

교수님 소개

석준희

연구실 소개

연구실 키워드

알럼나이가 간 회사

연구실 지원 방법

대우 조건

연도별 논문 인용수

저널 & 학회 논문 출판 수

연구실 논문 목록

Deriving Optimal Atomic Layer Deposition Process Conditions Using Machine Learning

Adversarial environment design for crowd navigation based on deep reinforcement learning

Automated BigSMILES conversion workflow and dataset for homopolymeric macromolecules

연구실 보유 특허

대학별 입시끈

우리 연구실 모집

교수님 소개

석준희

연구실 키워드

알럼나이가 간 회사

연구실 지원 방법

대우 조건

연도별 논문 인용수

저널 & 학회 논문 출판 수

연구실 논문 목록

Deriving Optimal Atomic Layer Deposition Process Conditions Using Machine Learning

Adversarial environment design for crowd navigation based on deep reinforcement learning

Automated BigSMILES conversion workflow and dataset for homopolymeric macromolecules

연구실 보유 특허

최근 3년간 연구실적 높은 연구실

대학별 입시끈

우리 연구실 모집