불균형 데이터 분류를 위한 딥러닝 기반 오버샘플링 기법

손민재,정승원,황인준 한국정보처리학회 2019 정보처리학회논문지. 소프트웨어 및 데이터 공학 Vol.8 No.7

분류 문제는 주어진 입력 데이터에 대해 해당 데이터의 클래스를 예측하는 문제로, 자주 쓰이는 방법 중의 하나는 주어진 데이터셋을 사용하여 기계학습 알고리즘을 학습시키는 것이다. 이런 경우 분류하고자 하는 클래스에 따른 데이터의 분포가 균일한 데이터셋이 이상적이지만, 불균형한 분포를 가지고 경우 제대로 분류하지 못하는 문제가 발생한다. 이러한 문제를 해결하기 위해 본 논문에서는 Conditional Generative Adversarial Networks(CGAN)을 활용하여 데이터 수의 균형을 맞추는 오버샘플링 기법을 제안한다. CGAN은 Generative Adversarial Networks(GAN)에서 파생된 생성 모델로, 데이터의 특징을 학습하여 실제 데이터와 유사한 데이터를 생성할 수 있다. 따라서 CGAN이 데이터 수가 적은 클래스의 데이터를 학습하고 생성함으로써 불균형한 클래스 비율을 맞추어 줄 수 있으며, 그에 따라 분류 성능을 높일 수 있다. 실제 수집된 데이터를 이용한 실험을 통해 CGAN을 활용한 오버샘플링 기법이 효과가 있음을 보이고 기존 오버샘플링 기법들과 비교하여 기존 기법들보다 우수함을 입증하였다. Classification problem is to predict the class to which an input data belongs. One of the most popular methods to do this is training a machine learning algorithm using the given dataset. In this case, the dataset should have a well-balanced class distribution for the best performance. However, when the dataset has an imbalanced class distribution, its classification performance could be very poor. To overcome this problem, we propose an over-sampling scheme that balances the number of data by using Conditional Generative Adversarial Networks (CGAN). CGAN is a generative model developed from Generative Adversarial Networks (GAN), which can learn data characteristics and generate data that is similar to real data. Therefore, CGAN can generate data of a class which has a small number of data so that the problem induced by imbalanced class distribution can be mitigated, and classification performance can be improved. Experiments using actual collected data show that the over-sampling technique using CGAN is effective and that it is superior to existing over-sampling techniques.

광학 영상의 구름 제거를 위한 조건부 생성적 적대 신경망과 회귀 기반 보정의 결합

곽근호,박소연,박노욱,Kwak, Geun-Ho,Park, Soyeon,Park, No-Wook 대한원격탐사학회 2022 大韓遠隔探査學會誌 Vol.38 No.6

Cloud removal is an essential image processing step for any task requiring time-series optical images, such as vegetation monitoring and change detection. This paper presents a two-stage cloud removal method that combines conditional generative adversarial networks (cGANs) with regression-based calibration to construct a cloud-free time-series optical image set. In the first stage, the cGANs generate initial prediction results using quantitative relationships between optical and synthetic aperture radar images. In the second stage, the relationships between the predicted results and the actual values in non-cloud areas are first quantified via random forest-based regression modeling and then used to calibrate the cGAN-based prediction results. The potential of the proposed method was evaluated from a cloud removal experiment using Sentinel-2 and COSMO-SkyMed images in the rice field cultivation area of Gimje. The cGAN model could effectively predict the reflectance values in the cloud-contaminated rice fields where severe changes in physical surface conditions happened. Moreover, the regression-based calibration in the second stage could improve the prediction accuracy, compared with a regression-based cloud removal method using a supplementary image that is temporally distant from the target image. These experimental results indicate that the proposed method can be effectively applied to restore cloud-contaminated areas when cloud-free optical images are unavailable for environmental monitoring.

차분 프라이버시를 만족하는 안전한 GAN 기반 재현 데이터 생성 기술 연구

강준영,정수용,홍도원,서창호 한국정보보호학회 2020 정보보호학회논문지 Vol.30 No.5

The publication of data is essential in order to receive high quality services from many applications. However, if theoriginal data is published as it is, there is a risk that sensitive information (political tendency, disease, ets.) may reveal. Therefore, many research have been proposed, not the original data but the synthetic data generating and publishing toprivacy preserve. but, there is a risk of privacy leakage still even if simply generate and publish the synthetic data byvarious attacks (linkage attack, inference attack, etc.). In this paper, we propose a synthetic data generation algorithm inwhich privacy preserved by applying differential privacy the latest privacy protection technique to GAN, which is drawingattention as a synthetic data generative model in order to prevent the leakage of such sensitive information. The generativemodel used CGAN for efficient learning of labeled data, and applied Rényi differential privacy, which is relaxation ofdifferential privacy, considering the utility aspects of the data. And validation of the utility of the generated data isconducted and compared through various classifiers. 많은 응용프로그램들로부터 양질의 서비스를 제공받기 위해서 데이터 공개는 필수적이다. 하지만 원본 데이터를그대로 공개할 경우 개인의 민감한 정보(정치적 성향, 질병 등)가 드러날 위험이 있기 때문에 원본 데이터가 아닌재현 데이터를 생성하여 공개함으로써 프라이버시를 보존하는 많은 연구들이 제안되어왔다. 그러나 단순히 재현 데이터를 생성하여 공개하는 것은 여러 공격들(연결공격, 추론공격 등)에 의해 여전히 프라이버시 유출 위험이 존재한다. 본 논문에서는 이러한 민감한 정보의 유출을 방지하기 위해, 재현 데이터 생성 모델로 주목받고 있는 GAN에최신 프라이버시 보호 기술인 차분 프라이버시를 적용하여 프라이버시가 보존되는 재현 데이터 생성 알고리즘을 제안한다. 생성 모델은 레이블이 있는 데이터의 효율적인 학습을 위해 CGAN을 사용하였고, 데이터의 유용성 측면을고려하여 기존 차분 프라이버시보다 프라이버시가 완화된 Rényi 차분 프라이버시를 적용하였다. 그리고 생성된 데이터의 유용성에 대한 검증을 다양한 분류기를 통해 실시하고 비교분석하였다.

Land Use and Land Cover Mapping from Kompsat-5 X-band Co-polarized Data Using Conditional Generative Adversarial Network

장재철,박경애 대한원격탐사학회 2022 大韓遠隔探査學會誌 Vol.38 No.1

Land use and land cover (LULC) mapping is an important factor in geospatial analysis. Although highly precise ground-based LULC monitoring is possible, it is time consuming and costly. Conversely, because the synthetic aperture radar (SAR) sensor is an all-weather sensor with high resolution, it could replace field-based LULC monitoring systems with low cost and less time requirement. Thus, LULC is one of the major areas in SAR applications. We developed a LULC model using only KOMPSAT-5 single co-polarized data and digital elevation model (DEM) data. Twelve HH-polarized images and 18 VV-polarized images were collected, and two HH-polarized images and four VV-polarized images were selected for the model testing. To train the LULC model, we applied the conditional generative adversarial network (cGAN) method. We used U-Net combined with the residual unit (ResUNet) model to generate the cGAN method. When analyzing the training history at 1732 epochs, the ResUNet model showed a maximum overall accuracy (OA) of 93.89 and a Kappa coefficient of 0.91. The model exhibited high performance in the test datasets with an OA greater than 90. The model accurately distinguished water body areas and showed lower accuracy in wetlands than in the other LULC types. The effect of the DEM on the accuracy of LULC was analyzed. When assessing the accuracy with respect to the incidence angle, owing to the radar shadow caused by the side-looking system of the SAR sensor, the OAtended to decrease as the incidence angle increased. This study is the first to use only KOMPSAT-5 single co-polarized data and deep learning methods to demonstrate the possibility of high-performance LULC monitoring. This study contributes to Earth surface monitoring and the development of deep learning approaches using the KOMPSAT-5 data.

Frontal Face Generation Algorithm from Multi-view Images Based on Generative Adversarial Network

허영진,김병규,Partha Pratim Roy 한국멀티미디어학회 2021 The journal of multimedia information system Vol.8 No.2

In a face, there is much information of person's identity. Because of this property, various tasks such as expression recognition, identity recognition and deepfake have been actively conducted. Most of them use the exact frontal view of the given face. However, various directions of the face can be observed rather than the exact frontal image in real situation. The profile (side view) lacks information when comparing with the frontal view image. Therefore, if we can generate the frontal face from other directions, we can obtain more information on the given face. In this paper, we propose a combined style model based the conditional generative adversarial network (cGAN) for generating the frontal face from multi-view images that consist of characteristics that not only includes the style around the face (hair and beard) but also detailed areas (eye, nose, and mouth).

A Study on the Performance Improvement of Machine LearningClassification Model by Resolving Imbalance in Financial Data: Focusing on Credit Card Accidents

류혜선,김상회,오경주 계명대학교 자연과학연구소 2023 Quantitative Bio-Science Vol.42 No.1

The advancement of innovative technologies such as artificial intelligence (AI) and big data has considerably increased in the amount of financial data. However, the quantity of abnormal data belonging to minority classes is relatively low, resulting in an exacerbation of data imbalance phenomenon. Data imbalance significantly impairs the performance of classifiers for minority classes. In this study, classification performance comparison experiments were conducted according to the data imbalance resolution methodology using credit card customer data, one of the financial data. Empirical analysis revealed that as the degree of data imbalance decreased, the performance of the classification model tended to improve, and applying the data imbalance resolution methodology led to performance enhancement. When SMOTETomek was applied, the deep neural network (DNN) showed the highest performance improvement, and the performance of the classification model could be improved using a conditional adversarial generative network (CGAN) for financial data. This study proposed solutions for data imbalance that could be applied to financial data, and the proposed methods contributed to the adoption and development of AI technology in the financial industry by making it more suitable for financial practices.

A Study on Fuzzy Searching Algorithm and Conditional-GAN for Crime Prediction System

Carmelita Afonso(카멜리타 아폰소),Han-Kyung, Yun(윤한경) 한국정보전자통신기술학회 2021 한국정보전자통신기술학회논문지 Vol.14 No.2

본 연구에서는 현재 발생한 범죄와 과거 유사 범죄의 기록을 조사하여 용의선상에 오른 자들과 전과자들를 비교분석하여 범인를 예측하는 시스템을 제안한다. 제안된 시스템은 용의자들과 전과자들의 안면을 비교하기 위하여 조건부 생성 적대 네트워크를 포함하는 퍼지 매칭으로 예상 범인을 선별하는 인공 지능 기반 알고리즘 범죄 예측 시스템(CPS)입니다. 유효성을 증명하기 위하여동 티모르. 범죄 기록의 데이터를 활용하였습니다. 구축 된 알고리즘은 증언을 바탕으로 몽타쥬를 작성하여 범죄 기록상의 전과자 안면과 비교됩니다. 제안 된 알고리즘과 CPS의 결과는 범죄를 처리하는 경찰관의 시간과 노력을 최소화될 뿐만 아니라 신속한 결과를 얻었으므로 유용하다는 것을 확인했습니다. 특히, 동 티므로와 같이 부족한 인적 자원과 예산으로 사회 안전망을 유지하는 것이 어려운 국가에 제안된 시스템의 적용은 미해결 범죄의 감소와 신속한 범죄 수사에 기여할 수 있다. In this study, artificial intelligence-based algorithms were proposed, which included a fuzzy search for matching suspects between current and historical crimes in order to obtain related cases in criminal history, as well as conditional generative adversarial networks for crime prediction system (CPS) using Timor-Leste as a case study. By comparing the data from the criminal records, the built algorithms transform witness descriptions in the form of sketches into realistic face images. The proposed algorithms and CPS"s findings confirmed that they are useful for rapidly reducing both the time and successful duties of police officers in dealing with crimes. Since it is difficult to maintain social safety nets with inadequate human resources and budgets, the proposed implemented system would significantly assist in improving the criminal investigation process in Timor-Leste.

조건부 적대적 생성 신경망을 이용한 OFDM 통신채널 모델링

고만재(Man-Jae Go),필웅후(XIONG-HOU BI),정태건(Tea-Geon Chung),김세연(Se-Yeon Kim),고학림(Hak-Lim Ko),조용호(Yong-Ho Cho) 한국정보통신학회 2020 한국정보통신학회 종합학술대회 논문집 Vol.24 No.1

딥러닝 기술을 이용한 중첩 지문의 분리에 대한 연구

이재형(Jae Hyeong Lee),이환용(Hwanyong Lee) 한국HCI학회 2021 한국HCI학회 학술대회 Vol.2021 No.1

기상레이더 관측자료와 딥러닝을 이용한 초단기 강우예측

김예린,홍성욱 한국기상학회 2023 한국기상학회 학술대회 논문집 Vol.2023 No.4