• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2023년도 춘계학술발표대회
• /
• pp.527-529
• /
• 2023

Generative Adversarial Networks (GANs) models have developed rapidly due to the emergence of various variation models and their wide applications. Despite many recent developments in GANs, mode collapse, and instability are still unresolved issues. To address these problems, we focused on the fact that a single GANs model itself cannot realize local failure during the training phase without external standards. This paper introduces a novel training process involving multiple GANs, inspired by auction mechanisms. During the training, auxiliary performance metrics for each GANs are determined by the others through the process of various auction methods.

• 한국컴퓨터정보학회:학술대회논문집
• /
• 한국컴퓨터정보학회 2023년도 제68차 하계학술대회논문집 31권2호
• /
• pp.713-714
• /
• 2023

본 논문에서는 사전학습모델을 활용한 수학학습 도구 자동 생성 시스템을 제안한다. 본 시스템은 사전학습모델을 활용하여 수학학습 도구를 교과과정 및 단원, 유형별로 다각화하여 자동 생성하고 사전학습모델을 자체 구축한 Dataset을 이용해 Fine-tuning하여 학생들에게 적절한 학습 도구와 적절치 않은 학습 도구를 분류하여 학습 도구의 품질을 높이었다. 본 시스템을 활용하여 학생들에게 양질의 수학학습 도구를 많은 양으로 제공해 줄 수 있는 초석을 다지었으며, 추후 AI 교과서와의 융합연구의 가능성도 열게 되었다.

• Journal of Information Processing Systems
• /
• 제19권5호
• /
• pp.641-651
• /
• 2023

Malicious hate speech and gender bias comments are common in online communities, causing social problems in our society. Gender bias and hate speech detection has been investigated. However, it is difficult because there are diverse ways to express them in words. To solve this problem, we attempted to detect malicious comments in a Korean hate speech dataset constructed in 2020. We explored bidirectional encoder representations from transformers (BERT)-based deep learning models utilizing hyperparameter tuning, data sampling, and logits ensembles with a label distribution. We evaluated our model in Kaggle competitions for gender bias, general bias, and hate speech detection. For gender bias detection, an F1-score of 0.7711 was achieved using an ensemble of the Soongsil-BERT and KcELECTRA models. The general bias task included the gender bias task, and the ensemble model achieved the best F1-score of 0.7166.

• 한국컴퓨터그래픽스학회논문지
• /
• 제24권3호
• /
• pp.61-71
• /
• 2018

RGB-D 카메라 촬영 영상에 대한 카메라 포즈 추정을 통하여 복원한 3차원 전역 공간의 점 집합으로부터 삼각형 메쉬를 생성할 때, 일반적으로 메쉬의 크기가 커질수록 3차원 모델의 품질 또한 향상된다. 하지만 어떤 한계를 넘어서 삼각형 메쉬의 해상도를 높일 경우, 메모리 요구량의 과도한 증가나 실시간 렌더링 성능저하 문제뿐만 아니라 RGB-D 센서의 정밀도 한계로 인한 접 집합 데이터의 노이즈에 민감해지는 문제가 발생한다. 본 논문에서는 실시간 응용에 적합한 3차원 모델 생성을 위하여 비교적 적은 크기의 삼각형 메쉬에 대하여 3차원 점 집합의 촬영 색상으로부터 고화질의 텍스쳐를 생성하는 기법을 제안한다. 특히 카메라 포즈 추정을 통하여 생성한 3차원 점 집합 공간과 2차원 텍스쳐 공간 간의 매핑 관계를 활용한 간단한 방법을 통하여 RGB-D 카메라 촬영 영상으로부터 복원한 3차원 모델에 대하여 효과적으로 텍스쳐를 생성할 수 있음을 보인다.

• 정보화연구
• /
• 제9권3호
• /
• pp.243-251
• /
• 2012

본 논문은 온톨로지 기반의 임무공간 개념 모델링 체계인 한국형 임무공간 개념모델 (CMMSK:The Conceptual Models of the Mission Space-Korea)을 제안한다. 모델링과 시뮬레이션을 이용하면 실제 군사 훈련을 실행할 때 발생하는 시간, 공간 그리고 경제적인 비용을 크게 줄일 수 있다. CMMS-K는 국방 모델과 시뮬레이션의 상호운용성과 재사용성을 향상시키기 위해 개발되고 있다. CMMS-K의 구조는 한국 국방 환경을 기반으로 기존 국방 개념 모델링 체계를 참조하여 생성되었다. CMMS-K의 주요 구성요소는 도메인 온톨로지, 임무공간 모델 기술 언어, 임무공간 모델링 도구, 그리고 CMMS-K 관리 시스템이다. CMMS-K 도메인 온톨로지는 개체 및 과제 온톨로지로 구성된다. 본 논문은 CMMS-K 도메인 온톨로지에 대해서 자세히 기술하고 제안된 방법의 적용 가능성을 예제를 이용하여 평가한다.

• Geomechanics and Engineering
• /
• 제36권5호
• /
• pp.489-509
• /
• 2024

Slope stability analysis and prediction are of critical importance to geotechnical engineers, given the severe consequences associated with slope failure. This research endeavors to forecast the factor of safety (FOS) for slopes through the implementation of six distinct ML techniques, including back propagation neural networks (BPNN), feed-forward neural networks (FFNN), Takagi-Sugeno fuzzy system (TSF), gene expression programming (GEP), and least-square support vector machine (Ls-SVM). 344 slope cases were analyzed, incorporating a variety of geometric and shear strength parameters measured through the PLAXIS software alongside several loss functions to assess the models' performance. The findings demonstrated that all models produced satisfactory results, with BPNN and GEP models proving to be the most precise, achieving an R² of 0.86 each and MAE and MAPE rates of 0.00012 and 0.00002 and 0.005 and 0.004, respectively. A Pearson correlation and residuals statistical analysis were carried out to examine the importance of each factor in the prediction, revealing that all considered geomechanical features are significantly relevant to slope stability. However, the parameters of friction angle and slope height were found to be the most and least significant, respectively. In addition, to aid in the FOS computation for engineering challenges, a graphical user interface (GUI) for the ML-based techniques was created.

• ETRI Journal
• /
• 제42권1호
• /
• pp.90-100
• /
• 2020

Named entity recognition (NER) continues to be an important task in natural language processing because it is featured as a subtask and/or subproblem in information extraction and machine translation. In Urdu language processing, it is a very difficult task. This paper proposes various deep recurrent neural network (DRNN) learning models with word embedding. Experimental results demonstrate that they improve upon current state-of-the-art NER approaches for Urdu. The DRRN models evaluated include forward and bidirectional extensions of the long short-term memory and back propagation through time approaches. The proposed models consider both language-dependent features, such as part-of-speech tags, and language-independent features, such as the "context windows" of words. The effectiveness of the DRNN models with word embedding for NER in Urdu is demonstrated using three datasets. The results reveal that the proposed approach significantly outperforms previous conditional random field and artificial neural network approaches. The best f-measure values achieved on the three benchmark datasets using the proposed deep learning approaches are 81.1%, 79.94%, and 63.21%, respectively.

• Industrial Engineering and Management Systems
• /
• 제12권1호
• /
• pp.23-29
• /
• 2013

This paper presents inventory models for fresh agriculture products with time-varying deterioration rate. Due to the particularity of fresh agriculture products, the demand rate is a function that depends on sale price and freshness. The deterioration rate increases with time and is assumed to be a time-varying function. In the models, the inventory cycle may be constant or variable. The optimal solutions of models are discussed for different freshness and the deterioration rate. The results of experiments show that the profit depends on the freshness and deterioration rate of products. With the increasing inventory cycle, the sale price and profit increase at first and then start decreasing. Furthermore, when the inventory cycle is variable, the total profit is a binary function of the sale price and inventory cycle. There exist unique sale price and inventory cycle such that the profit is optimal. The results also show that the optimal sale price and inventory cycle depend on the freshness and the deterioration rate of fresh agriculture products.

• 한국컴퓨터산업학회논문지
• /
• 제4권4호
• /
• pp.491-500
• /
• 2003

네트워크 보안은 정보통신 및 인터넷 기술이 발전함에 따라 그 중요성과 필요성이 더욱 절실해지고 있다. 본 연구에서는 침입차단 시스템, 운영체제 모델과 다양한 네트워크 구성요소들을 모델링 하였다. 각 모델은 MODSIM III 기반의 기본모델(Basic Model)과 결합모델(Compound Model)의 두 가지 유형으로 정의하였다. 대상 네트워크 환경에서 사용한 공격은 서비스 거부공격 형태인 SYN flooding 공격과 Smurf 공격을 발생하였다. 이 공격들에 대하여 패킷 필터 모델에 다양한 보안 정책을 적용하여 시뮬레이션을 실행하였다. 본 연구에서의 시뮬레이션을 통하여 보안정책의 강도를 점점 높였을 때 보안성능이 향상되는 점을 검증하였다.

• KIEE International Transactions on Power Engineering
• /
• 제4A권1호
• /
• pp.26-32
• /
• 2004

Digital technology has advanced significantly over the years both in terms of software tools and hardware availability. It is now applied extensively throughout many area of electrical engineering including protective relaying in power systems. Digital relays have numerous advantages over traditional analog relays, such as the ability to accomplish what is difficult or impossible using analog relays. Although non real-time simulators like PSCAD/EMTDC are employed to test the algorithms, such simulations are disadvantaged in that they cannot test the relay dynamically. Hence, real-time simulators like RTDS are used. However, the latter requires large space and is very expensive. This paper uses EMTP MODELS to simulate the power system and the distance relay. The distance relay algorithm is implemented and the distance relay is interfaced with a test power system. The distance relay's performance is then assessed interactively under various fault types, fault distances and fault inception angles. The test results show that we can simulate the distance relay effectively and we can examine the operation of the distance relay very closely including its drawbacks/limitations by using EMTP MODELS. Equally important, this approach facilitates any changes that need to be carried out in order to enhance the Distance Relay under test/examination.