• KIPS Transactions on Software and Data Engineering
• /
• v.6 no.12
• /
• pp.573-580
• /
• 2017

Deep Neural Network (DNN) is a large layered neural network which is consisted of a number of layers of non-linear units. Deep Learning which represented as DNN has been applied very successfully in various applications. However, many issues in DNN have been identified through past researches. Among these issues, generalization is the most well-known problem. A Recent study, Dropout, successfully addressed this problem. Also, Dropout plays a role as noise, and so it helps to learn robust feature during learning in DNN such as Denoising AutoEncoder. However, because of a large computations required in Dropout, training takes a lot of time. Since Dropout keeps changing an inter-layer representation during the training session, the learning rates should be small, which makes training time longer. In this paper, using mutation operation, we reduce computation and improve generalization performance compared with Dropout. Also, we experimented proposed method to compare with Dropout method and showed that our method is superior to the Dropout one.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2011.06a
• /
• pp.175-177
• /
• 2011

본 논문의 Live Site 개념을 도입한 웹사이트 저작도구의 실시간 자동화 모델에 관한 연구에서는 MVC(Model-View-Controller) 패턴의 시스템으로 사용자가 요구 사항을 전달하면 jsp에서 jsp 간의 호출을 통해 서버로 변경 사항을 넘기고 그 결과물을 다시 사용자에게 보여주는 형식으로 진행된다. 본 시스템 개발에 사용한 Jquery는 자바스크립트와 HTML 사이의 상호작용을 강조하는 경량화된 web application framework로 일반적 웹 스크립팅에 폭넓게 사용 될 수 있는 추상적 계층을 제공하여 스크립팅에서 필요로 하는 거의 모든 상황에 사용 할 수 있다. 본 논문에서는 추상화된 데이터를 제공하여 일상적인 작업들을 일반화 하고 코드의 크기를 줄이며 극도로 단순하게 개발이 가능한 jquery를 사용하여 거의 모든 브라우저에 호환이 가능한, 사용자 각 개인의 경향에 맞춘 웹사이트 저작 도구를 개발하였다. 본 논문에서는 추상화된 데이터를 제공하므로 일상적인 작업들을 일반화 하고 코드의 크기를 줄이며 극도로 단순하게 개발이 가능한 jquery를 사용하여 거의 모든 브라우저에 호환이 가능한, 사용자 각 개인의 경향에 맞춘 웹 저작 도구를 연구하였다. 전 세계적으로 웹 시장이 대두 되는 이 시점에 본 프로그램은 다양한 웹 제작 공급에 대한 새로운 시장을 형성해 주며, 새로운 콘텐츠 제작 방식의 도입으로 인한 활발한 인터넷 시장이 형성 되리라 기대한다. 현재 일부 생소한 Live Site 개념 즉, '사용자가 직접 보고 느끼며 원하는 대로 만드는 웹' 이란 개념의 가지고 고객 만족 커뮤니티라는 목적에 중점을 둔 본 프로그램 개발은 최근 웹 경향에 따른 이상적인 시스템이라 할 수 있다.

• The Journal of the Convergence on Culture Technology
• /
• v.10 no.1
• /
• pp.435-441
• /
• 2024

In order for the large amount of collected data sets to be used as deep learning training data, sensitive personal information such as resident registration number and disease information must be changed or encrypted to prevent it from being exposed to hackers, and the data must be reconstructed to match the structure of the built deep learning model. Currently, these tasks are performed manually by experts, which takes a lot of time and money. To solve these problems, this paper proposes a technique that can automatically perform data processing tasks to protect personal information during the deep learning process. In the proposed technique, privacy protection tasks are performed based on data generalization and data reconstruction tasks are performed using circular queues. To verify the validity of the proposed technique, it was directly implemented using C language. As a result of the verification, it was confirmed that data generalization was performed normally and data reconstruction suitable for the deep learning model was performed properly.

• Journal of KIISE:Software and Applications
• /
• v.37 no.2
• /
• pp.120-128
• /
• 2010

A hypernetwork is a generalized hypo-graph and a probabilistic graphical model based on evolutionary learning. Hypernetwork models have been applied to various domains including pattern recognition and bioinformatics. Nevertheless, conventional hypernetwork models have the limitation that they can manage data with categorical or discrete attibutes only since the learning method of hypernetworks is based on equality comparison of hyperedges with learned data. Therefore, real-valued data need to be discretized by preprocessing before learning with hypernetworks. However, discretization causes inevitable information loss and possible decrease of accuracy in pattern classification. To overcome this weakness, we propose a novel feature-wise L1-distance based method for real-valued attributes in learning hypernetwork models in this study. We show that the proposed model improves the classification accuracy compared with conventional hypernetworks and it shows competitive performance over other machine learning methods.

• Journal of the Earthquake Engineering Society of Korea
• /
• v.10 no.2 s.48
• /
• pp.1-10
• /
• 2006

Reinforced concrete bridge columns with relatively small aspect ratio show flexure-shear behavior, which is flexural behavior at initial and medium displacement stages and shear failure at final stage. Since the columns with flexure-shear failure have lower ductility than those with flexural failure, shear capacity curve models shall be applied as well as flexural capacity curve in order to determine ultimate displacement for seismic design or performance evaluation. In this paper, a modified shear capacity curve model is proposed and compared with the other models such as the CALTRANS model, Aschheim et al.'s model, and Priestley et al.'s model. Four shear capacity curve models are applied to the 4 full scale circular bridge column test results and the accuracy of each model is discussed. It may not be fully adequate to drive a final decision from the application to the limited number of test results, however the proposed model provides the better prediction of failure mode and ultimate displacement than the other models for the selected column test results.

• Journal of KIISE:Software and Applications
• /
• v.30 no.12
• /
• pp.1172-1184
• /
• 2003

At present, new techniques based on different component reference models for the integration of component and system of different platforms, such as EJB and COM+, are introduced. The operation between the components in the identical component platform is realized by the composition at the source level. In case of the different component platform, however, it is impossible to use combined components in real condition although they are components of similar domain. In this paper we proposed a solution for the composition problem by using component transformation methodology based on model between EJB and COM+ components which are different components. For the composition between EJB and COM+ components, we compared and analyzed each reference model, then proposed the Virtual Component Model which is implementation independent and the Implementation Table for the mutual conversion. Reffering to the Virtual Component Model and the Implementation Table, we can generalize each Implementation model to the Virtual Component Model, make the Virtual Component Model which is implementation independent through the virtual component modeling, transform EJB and COM+ components selectively. Proposing the effective Model Transformation method to the different component platform, we can combine EJB and COM+ components.

• Journal of The Korean Association of Information Education
• /
• v.20 no.6
• /
• pp.645-654
• /
• 2016

As the 4th industrial revolution has progressed in recent years, the importance of robot education in elementary school education is increasing. In this paper, I suggested robot education framework to consider conceptual understanding and learning activities based on the 2014, 2015 KAIE software education standard curriculum for elementary school. The framework is reconstructed the 7 stages, In order to generalize the standardized model of the software curriculum, the achievement criteria should be prepared according to the content system of the curriculum considering the conceptual understanding and learning activities proposed in this paper, and if the educational contents are developed and utilized, it is expected to contribute to the activation of robot education in addition to elementary school software education.

• Proceedings of the KIEE Conference
• /
• 2008.10b
• /
• pp.247-248
• /
• 2008

본 논문에서는 TYPE-2 퍼지 추론 기반의 RBF 뉴럴 네트워크(TYPE-2 Radial Basis Function Neural Network, T2RBFNN)를 설계하고 PSO(Particle Swarm Optimization) 알고리즘을 이용하여 모델의 파라미터를 동정한다. 제안된 모델의 은닉층은 TYPE-2 가우시안 활성 함수로 구성되며, 출력층은 Interval set 형태의 연결가중치를 갖는다. 여기에서 규칙 전반부 활성함수의 중심 선택은 C-means 클러스터링 알고리즘을 이용하고, 규칙 후반부 Interval set 형태의 연결가중치 결정에는 경사 하강법(Gradient descent method)을 이용한 오류 역전파 알고리즘을 사용하여 학습한다. 또한, 최적의 모델을 설계하기 위한 학습율 및 활성함수의 활성화 영역 결정에는 입자 군집 최적화(PSO; Particle Swarm Optimization) 알고리즘으로 동조한다. 마지막으로, 제안된 모델의 평가를 위하여 모의 데이터 집합(Synthetic dadaset)을 적용하고 근사화 및 일반화 능력에 대하여 토의한다.

• Journal of KIISE:Software and Applications
• /
• v.30 no.7_8
• /
• pp.744-751
• /
• 2003

A hybrid least square Support Vector Machine combined with First Principle(FP) knowledge is proposed. We compare hybrid least square Support Vector Machine(HLS-SVM) with early proposed models such as Hybrid Neural Network(HNN) and HNN with Extended Kalman Filter(HNN-EKF). In the training and validation stage HLS-SVM shows similar performance with HNN-EKF but better than HNN, whereas, in the testing stage, it shows three times better than HNN-EKF, hundred times better than HNN model.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.13 no.2
• /
• pp.149-155
• /
• 2013

In these days, Entity-Relationship Model is the most popular modeling tool for designing databases, and XML is a de facto standard language for representing and exchanging data. But, because of many commercial products supporting Entity-Relationship Model use their's own representation formats, and thus it gives rise to difficulties the inter-operability between these products. In this paper, we propose an ERX, a generation tool of XML Schema from Entity-Relationship Model. ERX receives an Entity-Relationship Diagram as an input, transforms it based on transformation rules, and generates a XML Schema Definition as an output. Transformation rules contain entity set, relationship set, mapping cardinalities, and generalization.