• 방송공학회논문지
• /
• 제26권7호
• /
• pp.844-854
• /
• 2021

Semi-supervised learning (SSL) and few-shot learning (FSL) have shown impressive performance even then the volume of labeled data is very limited. However, SSL and FSL can encounter a significant performance degradation if the diversity gap between the labeled and unlabeled data is high. To reduce this diversity gap, we propose a novel scheme that relies on an autoencoder for generating pseudo examples. Specifically, the autoencoder is trained on a specific class using the available labeled data and the decoder of the trained autoencoder is then used to generate N samples of that specific class based on N random noise, sampled from a standard normal distribution. The above process is repeated for all the classes. Consequently, the generated data reduces the diversity gap and enhances the model performance. Extensive experiments on MNIST and FashionMNIST datasets for SSL and FSL verify the effectiveness of the proposed approach in terms of classification accuracy and robustness against adversarial attacks.

• 한국정보시스템학회지:정보시스템연구
• /
• 제5권
• /
• pp.149-173
• /
• 1996

This paper focus on the development of object-oriented model bases for Structured Modeling. For the model base organization, object modeling techniques and model typing concept which is similar to data typing concept are used. Structured modeling formalizes the notion of a definitional system as a way of dscribing models. From the object-oriented concept, a structured model can be represented as follows. Each group of similar elements(genus) is represented by a composite class. Other type of genera can be represented in a similar manner. This hierarchical class composition gives rise to an acyclic class-composition graph which corresponds with the genus graph of structured model. Nodes in this graph are instantiated to represent the elemental graph for a specific model. Taking this class composition process one step further, we aggregate the classes into higher-level composite classes which would correspond to the structured modeling notion of a module. Finally, the model itself is then represented by a composite class having attributes each of whose domain is a composite class representing one of the modules. The resulting class-composition graph represent the modular tree of the structured.

• Journal of the Korean Statistical Society
• /
• 제27권4호
• /
• pp.451-458
• /
• 1998

Problem of testing linearity among general class of first order nonlinear time series models is discussed. The null hypotheses of linearity is identified via conditional expectations. A consistent test is then suggested and relevant limiting results are derived. It is worth indicating that any specific alternatives are not specified.

• Bulletin of the Korean Chemical Society
• /
• 제32권10호
• /
• pp.3756-3759
• /
• 2011

• 한국운동역학회지
• /
• 제24권3호
• /
• pp.209-216
• /
• 2014

The purpose of this study was to explore the effect of 8 weeks after school classes of track sports on students' start motion through kinematic variables. 30 students in D National university of education participated for this study. These students divided into two groups, 17 students for experimental group and 13 students for control group. The two groups participated in general athletic class as common class, and the experimental group participated in after school class additionally. The general class taught track and jump skill for 3 hours a week, and the after school class taught only athletic running skill for 30 minutes a week. Pre and post test assessed to assess students' kinematic changes. Findings indicated that velocity and step rate were increased, and contact time, step length, displacement of center of mass were decreased in the experimental group. In the control group, early velocity, knee/hip angle velocity were increased, and contact time, step length, displacement of center of mass were decreased. In conclusion, the students who only participated in general class, could not maintain athletic skills obtained from the class. However, students who participated in both general and after school class accomplish athletic start skill. This is because Continuous learning effect helped students keep the skill and did not lose the skill. Thus, in order for students to learn specific sport skills, joining after school class with general class together is recommended.

• 정보처리학회논문지:소프트웨어 및 데이터공학
• /
• 제4권6호
• /
• pp.247-252
• /
• 2015

소프트웨어의 규모가 커지고 복잡성이 증가함에 따라 소프트웨어의 유지보수가 보다 중요해지고 있으며 유지보수성에 많은 영향을 미치는 요인 중 하나는 소스코드 가독성이다. 가독성의 90% 이상 영향을 끼치는 요인은 소스코드에서 사용되는 식별자들의 이름이며 이를 위한 기존 연구들에서는 클래스의 식별자로 사용된 어휘를 이용하여 식별자의 이름을 검증한다. 하지만 대부분의 관련 연구는 그 특성상 개체의 도메인 관련 특성만을 고려하게 되며 클래스 내의 어휘가 적절하지 못한 경우 적용할 수 있는 범위가 한정적이라는 한계점이 있다. 본 논문에서는 클래스의 특성을 추출하여 의사결정트리 기법을 통해 기계학습을 시킨 후 클래스 역할 모델을 생성하며 이를 이용하여 이름을 검증할 대상 클래스의 역할에 해당하는 접미사를 추천하게 되어 클래스 이름 검증 보고서를 생성한다. 본 연구 기법의 효용성을 검증하기 위해 4개의 오픈소스 프로젝트에 대하여 본 연구 기법을 적용하였고 클래스 역할 정보를 담고 있는 5개의 접미사에 대해 정확도와 재현율, ROC 곡선과 같은 지표를 제시하였다.

• 대한수학회보
• /
• 제40권2호
• /
• pp.243-251
• /
• 2003

This note Presents sufficient conditions for Lyapunov's stability and instability of a class of nonlinear nonautonomous second-order ordinary differential equations. Such a class includes as particular cases a remarkably large number of differential equations with specific physical applications. Two successive nonlinear transformations are applied to the original differential equation in order to convert it into a more convenient form for stability analysis purposes. The obtained stability / instability conditions depend closely on the parameterization of the original differential equation.

• Communications for Statistical Applications and Methods
• /
• 제6권2호
• /
• pp.511-522
• /
• 1999

The paper considers the problem of robustness in hierarchical bayesian models. In specific we address Bayesian robustness in the estimation of normal means. We provide the ranges of the posterior means under $\varepsilon$-contamination class as well as the density ratio class of priors. For the class of priors that are uniform over a specified interval we investigate the sensitivity as to the choice of the intervals. The methods are illustrated using the famous baseball data of Efron and Morris(1975).

• 한국경영과학회지
• /
• 제18권2호
• /
• pp.147-156
• /
• 1993

Recently a new concept of shadow prices, called average shadow price, has been developed. This paper provides a dual problem and the corresponding duality theorems justifying this new shadow price. The general duality framework is used. As an important secondary result, a new reduced class of price function, the pp. h.-class, has been developed for the general duality theory. This should be distinguished from other known reductions achieved in some specific areas of mathematical programming, in that it sustains the strong duality property in all the mathematical programs. The new general dual problem suggested with this pp. h.-class provides, as an optimal solution, the average shadow prices.

• International Journal of Computer Science & Network Security
• /
• 제23권9호
• /
• pp.129-133
• /
• 2023

Classification or prediction problem is how to solve it using a specific feature to obtain the predicted class. A wheat seeds specifications 4 3 classes of seeds will be used in a prediction process. A multi linear regression will be built, and a prediction error ratio will be calculated. To enhance the prediction ratio an ANN model will be built and trained. The obtained results will be examined to show how to make a prediction tool capable to compute a predicted class number very close to the target class number.