• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2011년도 춘계학술발표대회
• /
• pp.78-80
• /
• 2011

In the area of data fusion, dealing with heterogeneous data sources, numerous models have been proposed in last three decades to facilitate different application domains i.e. Department of Defense (DoD), monitoring of complex machinery, medical diagnosis and smart buildings. All of these models shared the theme of multiple levels processing to get more reliable and accurate information. In this paper, we consider five most widely acceptable fusion models (Intelligence Cycle, Joint Directors of Laboratories, Boyd control, Waterfall, Omnibus) applied to different areas for data fusion. When they are exposed to a real scenario, where large dataset from heterogeneous sources is utilize for object monitoring, then it may leads us to non-efficient and unreliable information for decision making. The proposed variation works better in terms of time and accuracy due to prior data diminution.

• 한국정보과학회 언어공학연구회:학술대회논문집(한글 및 한국어 정보처리)
• /
• 한국정보과학회언어공학연구회 2023년도 제35회 한글 및 한국어 정보처리 학술대회
• /
• pp.196-202
• /
• 2023

Large Language Models (LLM)의 출현은 자연어 처리 분야의 연구 패러다임을 전환시켰다. LLM의 핵심적인 성능향상은 지시어 튜닝(instruction-tuning) 기법의 결과로 알려져 있다. 그러나, 현재 대부분의 연구가 영어 중심으로 진행되고 있어, 다양한 언어에 대한 접근이 필요하다. 본 연구는 한국어 지시어(instruction-following) 모델의 개발 및 최적화 방법을 제시한다. 본 연구에서는 한국어 지시어 데이터셋을 활용하여 LLM 모델을 튜닝하며, 다양한 데이터셋 조합의 효과에 대한 성능 분석을 수행한다. 최종 결과로 개발된 한국어 지시어 모델을 오픈소스로 제공하여 한국어 LLM 연구의 발전에 기여하고자 한다.

• Journal of the Korean Data and Information Science Society
• /
• 제13권1호
• /
• pp.25-34
• /
• 2002

In this paper, we proposed multiple trial unrelated question models that more efficient by reducing the variance of the estimate than single trial unrelated question models investigated by Greenberg et al.'s (1969) and Kim et al.'s (1992) an d Lee & Hong's (1998).

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2001년도 ICCAS
• /
• pp.156.6-156
• /
• 2001

A computer software system which enables to assess the air combat performance only by a computer is currently under development. The system is composed with plural aircraft models, missile models, bullet models etc. The aircraft can implement several empirical air combat maneuvers automatically depending on the situation , therefore air combat simulations and assessment can be attained. Some of these maneuvers and features are explained.

• International Journal of Advanced Culture Technology
• /
• 제11권3호
• /
• pp.310-314
• /
• 2023

The purpose of this study is to predict the remaining capacity of lithium-ion batteries and evaluate their performance using five artificial intelligence models, including linear regression analysis, decision tree, random forest, neural network, and ensemble model. We is in the study, measured Excel data from the CS2 lithium-ion battery was used, and the prediction accuracy of the model was measured using evaluation indicators such as mean square error, mean absolute error, coefficient of determination, and root mean square error. As a result of this study, the Root Mean Square Error(RMSE) of the linear regression model was 0.045, the decision tree model was 0.038, the random forest model was 0.034, the neural network model was 0.032, and the ensemble model was 0.030. The ensemble model had the best prediction performance, with the neural network model taking second place. The decision tree model and random forest model also performed quite well, and the linear regression model showed poor prediction performance compared to other models. Therefore, through this study, ensemble models and neural network models are most suitable for predicting the remaining capacity of lithium-ion batteries, and decision tree and random forest models also showed good performance. Linear regression models showed relatively poor predictive performance. Therefore, it was concluded that it is appropriate to prioritize ensemble models and neural network models in order to improve the efficiency of battery management and energy systems.

• 한국컴퓨터정보학회논문지
• /
• 제28권10호
• /
• pp.67-76
• /
• 2023

딥러닝 모델이 컴퓨터 비전 분야에서 혁신적인 성과를 이루어내고 있으나, 적대적 예제에 취약하다는 문제가 지속적으로 제기되고 있다. 적대적 예제는 이미지에 미세한 노이즈를 주입하여 오분류를 유도하는 공격 방법으로서, 현실 세계에서의 딥러닝 모델 적용에 심각한 위협이 될 수 있다. 본 논문에서는 객체의 엣지를 강조하여 학습된 분류 모델과 기본 분류 모델 간 예측 값의 차이를 이용하여 적대적 예제를 탐지하는 모델을 제안한다. 객체의 엣지를 추출하여 학습에 반영하는 과정만으로 분류 모델의 강건성을 높일 수 있으며, 모델 간 예측값의 차이를 통하여 적대적 예제를 탐지하기 때문에 경제적이면서 효율적인 탐지가 가능하다. 실험 결과, 적대적 예제(eps={0.02, 0.05, 0.1, 0.2, 0.3})에 대한 일반 모델의 분류 정확도는 {49.9%, 29.84%, 18.46%, 4.95%, 3.36%}를 보인 반면, Canny 엣지 모델은 {82.58%, 65.96%, 46.71%, 24.94%, 13.41%}의 정확도를 보였고 다른 엣지 모델들도 이와 비슷한 수준의 정확도를 보여, 엣지 모델이 적대적 예제에 더 강건함을 확인할 수 있었다. 또한 모델 간 예측값의 차이를 이용한 적대적 예제 탐지 결과, 각 epsilon별 적대적 예제에 대하여 {85.47%, 84.64%, 91.44%, 95.47%, 87.61%}의 탐지율을 확인할 수 있었다. 본 연구가 관련 연구 분야 및 의료, 자율주행, 보안, 국방 등의 응용 산업 분야에서 딥러닝 모델의 신뢰성 제고에 기여할 것으로 기대한다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제7권12호
• /
• pp.3274-3292
• /
• 2013

Unsupervised methods for image segmentation are recently drawing attention because most images do not have labels or tags. A topic model is such an unsupervised probabilistic method that captures latent aspects of data, where each latent aspect, or a topic, is associated with one homogeneous region. The results of topic models, however, usually have noises, which decreases the overall segmentation performance. In this paper, to improve the performance of image segmentation using topic models, we propose two topic masks applicable to topic assignments of homogeneous regions obtained from topic models. The topic masks capture the noises among the assigned topic assignments or topic labels, and remove the noises by replacements, just like image masks for pixels. However, as the nature of topic assignments is different from image pixels, the topic masks have properties that are different from the existing image masks for pixels. There are two contributions of this paper. First, the topic masks can be used to reduce the noises of topic assignments obtained from topic models for image segmentation tasks. Second, we test the effectiveness of the topic masks by applying them to segmented images obtained from the Latent Dirichlet Allocation model and the Spatial Latent Dirichlet Allocation model upon the MSRC image dataset. The empirical results show that one of the masks successfully reduces the topic noises.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제11권2호
• /
• pp.1217-1233
• /
• 2017

Effective monitoring and testing during each step are essential for document verification in research and development (R&D) projects. In software development, proper testing is required to verify it carefully and constantly because of the invisibility features of software. However, not enough studies on test processes for R&D projects have been done. Thus, in this paper, we introduce a Test Maturity Model integration (TMMi)-based software field R&D test process that offers five integrity levels and makes the process compatible for different types of projects. The Software & Systems Process Engineering Metamodel (SPEM) is used widely in the software process-modeling context, but it lacks built-in enactment capabilities, so there is no tool or process engine that enables one to execute the process models described in SPEM. Business Process Model and Notation (BPMN)-based workflow engines can be a solution for process execution, but process models described in SPEM need to be converted to BPMN models. Thus, we propose an approach to support enactment of SPEM-based process models by converting them into business processes. We show the effectiveness of our approach through converting software R&D test processes specified in SPEM in a case study.

• International Journal of Automotive Technology
• /
• 제7권3호
• /
• pp.309-314
• /
• 2006

In this paper, the ride comfort from a computer simulation was compared to the experimental result. For measuring ride comfort of a passenger car, acceleration data was obtained from the floor and seat during highway running with different speeds. The measured acceleration components were multiplied by the proper weighting functions, and then summed together to calculate overall ride values. Testing several passenger cars, the ride comforts were compared. In order to investigate the effect of vibration signals on the steering wheel, an apparatus to measure the vibrations and weighting functions on the steering wheel were designed. The effect of the steering accelerations on the ride comfort were investigated and added for the overall ride comfort. For the computer simulations, Korean dummy models were developed based on the Hybrid III dummy models. For the Korean dummy scaling, the national anthropometric survey of Korean people was used. In order to compare and check the validity of the developed Korean dummy models, dynamic responses were compared to those of Hybrid III dummy models. The computer simulation using the MADYMO software was also compared to the experimental results.

• International Journal of Computer Science & Network Security
• /
• 제21권12spc호
• /
• pp.526-538
• /
• 2021

Machine and deep learning-based models are emerging techniques that are being used to address prediction problems in biomedical data analysis. DNA sequence prediction is a critical problem that has attracted a great deal of attention in the biomedical domain. Machine and deep learning-based models have been shown to provide more accurate results when compared to conventional regression-based models. The prediction of the gene sequence that leads to cancerous diseases, such as prostate cancer, is crucial. Identifying the most important features in a gene sequence is a challenging task. Extracting the components of the gene sequence that can provide an insight into the types of mutation in the gene is of great importance as it will lead to effective drug design and the promotion of the new concept of personalised medicine. In this work, we extracted the exons in the prostate gene sequences that were used in the experiment. We built a Deep Neural Network (DNN) and Bi-directional Long-Short Term Memory (Bi-LSTM) model using a k-mer encoding for the DNA sequence and one-hot encoding for the class label. The models were evaluated using different classification metrics. Our experimental results show that DNN model prediction offers a training accuracy of 99 percent and validation accuracy of 96 percent. The bi-LSTM model also has a training accuracy of 95 percent and validation accuracy of 91 percent.