• International Journal of Advanced Culture Technology
• /
• 제9권3호
• /
• pp.188-201
• /
• 2021

The purpose of this study is to analyze the effects of online home learning in connection with extracurricular activities for lifelong education for the disabled in university on the cooking aids skills of cafeterias for the job search persons with developmental disabilities. Three people with job search developmental disabilities who have been in a state of unemployment for three years after graduating from a special high school course participated in the experiment. In order to verify the meaningful functional relationship between independent variables and dependent variables, multiple probe design across subjects, one of the main techniques of a single object study, was used. The experimental conditions according to the research design consisted of the steps of baseline, intervention, maintenance, and generalization. The dependent variable of this study is the restaurant cooking aid skills in the cafeteria, and three subskills such as side dish arrangement, sink arrangement, and dish washing were combined by task analysis. And the independent variable of this study was composed of procedures and methods to teach the environment, tools and materials related to the performance of dependent variables to the developmental disabled people at home by using real-time image technique through zoom service, and the contents of the performance by stages of task analysis. In addition, independent variables were applied to the subjects in the course of the extracurricular activities with the theme and contents of lifelong education for the disabled at university. Students who completed the above extracurricular activities practiced the intervention scene of the researcher through the screen sharing of zoom service. As a result, the subjects with developmental disabilities effectively acquired and maintained the positive response performance of dependent variables through independent variables. The subjects also showed high positive responses to generalization tests conducted in kitchens in cafeterias located elsewhere in the same university.

• 반도체디스플레이기술학회지
• /
• 제22권1호
• /
• pp.11-17
• /
• 2023

This paper proposes a lightweight deep learning network for detecting an image splicing forgery. The research on image forgery detection using CNN, a deep learning network, and research on detecting and localizing forgery in pixel units are in progress. Among them, CAT-Net, which learns the discrete cosine transform coefficients of images together with images, was released in 2022. The DCT coefficients presented by CAT-Net are combined with the JPEG artifact learning module and the backbone model as pre-learning, and the weights are fixed. The dataset used for pre-training is not included in the public dataset, and the backbone model has a relatively large number of network parameters, which causes overfitting in a small dataset, hindering generalization performance. In this paper, this learning module is designed to learn the characterization depending on the DCT domain in real-time during network training without pre-training. The DCT RRU-Net proposed in this paper is a network that combines RRU-Net which detects forgery by learning only images and JPEG artifact learning module. It is confirmed that the network parameters are less than those of CAT-Net, the detection performance of forgery is better than that of RRU-Net, and the generalization performance for various datasets improves through the network architecture and training method of DCT RRU-Net.

• 정보처리학회논문지:소프트웨어 및 데이터공학
• /
• 제6권12호
• /
• pp.573-580
• /
• 2017

심층 신경망은 많은 노드의 층을 쌓아 만든 거대한 신경망이다. 심층 신경망으로 대표되는 딥 러닝은 오늘날 많은 응용 분야에서 괄목할만한 성과를 거두고 있다. 하지만 다년간의 연구를 통해 심층 신경망에 대한 다양한 문제점이 식별되고 있다. 이 중 일반화는 가장 널리 알려진 문제점들 중 하나이며, 최근 연구 결과인 드롭아웃은 이러한 문제를 어느 정도 성공적으로 해결하였다. 드롭아웃은 노이즈와 같은 역할을 하여 신경망이 노이즈에 강건한 데이터 표현형을 학습할 수 있도록 하는데, 오토인코더와 관련된 연구에서 이러한 효과가 입증되었다. 하지만 드롭아웃은 빈번한 난수 연산과 확률연산으로 인해 신경망의 학습 시간이 길어지고, 신경망 각 계층의 데이터 분포가 크게 변화하여 작은 학습율을 사용해야하는 단점이 있다. 본 논문에서는 돌연변이 연산을 사용하여 비교적 적은 횟수의 연산으로 드롭아웃과 동등 이상의 성능을 나타내는 모델을 제시하고, 실험을 통하여 논문에서 제시한 방법이 드롭아웃 방식과 동등한 성능을 보임과 동시에 학습 시간 문제를 개선함을 보인다.

• 전기학회논문지
• /
• 제57권11호
• /
• pp.2095-2101
• /
• 2008

In this paper, we consider a performance improvement of neural network for predicting defect size of steam generator tube using early stopping. Usually, neural network is trained until MSE becomes less than a prescribed error goal. The smaller the error goal, the greater the prediction performance for the trained data. However, as the error goal is decreased, an over fitting is likely to start during supervised training of a neural network, which usually deteriorates the generalization performance. We propose that, for the prediction of an axisymmetric defect size, early stopping can be used to avoid the over-fitting. Through various experiments on the axisymmetric defect samples, we found that the difference bet ween the prediction error of neural network based on early stopping and that of ideal neural network is reasonably small. This indicates that the error goal used for neural network training for the prediction of defect size can be efficiently selected by early stopping.

• 설비공학논문집
• /
• 제24권6호
• /
• pp.538-544
• /
• 2012

Analysis of thermal performance is required for the economic operation of turbine cycle of power plant. We developed corrective model of main feed water flow which is the most important parameter for the precise analysis of turbine cycle performance. Classification model for the identification of feed water flow measurement status was applied to increase the suitability of the corrective model. We used neural network and support vector machine to develop estimation model of main feed water flow with more generalization capability. The estimation model can be used practically to evaluate corrective performance of turbine cycle plant.

• Journal of the Korean Data and Information Science Society
• /
• 제13권2호
• /
• pp.55-64
• /
• 2002

In this paper, we introduce SVM(support vector machine) developed to solve the problem of generalization of neural networks. We also introduce boosting algorithm which is a general method to improve accuracy of some given learning algorithm. We propose a new algorithm combining SVM and boosting to solve classification problem. Through the experiment with real and simulated data sets, we can obtain better performance of the proposed algorithm.

• Journal of the Korean Data and Information Science Society
• /
• 제19권1호
• /
• pp.25-36
• /
• 2008

Error Correcting Output Codes (ECOC) can improve generalization performance when applied to multicategory classification problem. In this study we propose a new criterion to select hyperparameters included in ECOC scheme. Instead of margins of a data we propose to use the probability of misclassification error since it makes the criterion simple. Using this we obtain an upper bound of leave-one-out error of OVA(one vs all) method. Our experiments from real and synthetic data indicate that the bound leads to good estimates of parameters.

• 한국통신학회논문지
• /
• 제26권8B호
• /
• pp.1078-1084
• /
• 2001

In this paper, fuzzy-ARTMAP neural network is applied for compensating the nonlinearity of satellite communication channel. The fuzzy-ARTMAP is made of using fuzzy logic and ART neural network. By a match tracking process with vigilance parameter, fuzzy ARTMAP neural network achieves a minimax learning rule that minimizes predictive error and maximizes generalization. Thus, the system automatically learns a minimal number of recognition categories, or hidden units, to meet accuracy criteria. Simulation studies are performed over satellite nonlinear channels. The performance of proposed fuzzy-ARTMAP equalizer is compared with MLP-basis equalizers.

• 품질경영학회지
• /
• 제26권2호
• /
• pp.17-26
• /
• 1998

Taguchi's parameter design is a method for quality improvement by making the performance fo a system robust to noise. Parameter design with dynamic characteristics has been recently the subject of much interest. This paper is concerned with a review and a generalization of the Signal-to-Noise (SN) ratio, a quality measure for parameter design with dynamic characteristics, proposed by Taguchi. We present a method for determination and analysis of the generalized SN ratio and illustrate its implementation by example.

• 한국측량학회:학술대회논문집
• /
• 한국측량학회 2007년도 춘계학술발표회 논문집
• /
• pp.153-156
• /
• 2007

There have been many studies on extracting building boundaries from LiDAR(Light Detection And Ranging) data. In such studies, points are first segmented, then are further processed to get straight boundary lines that better approximate the real boundaries. In most research in this area, processes like generalization or regularization assume that buildings have only right angles, i.e. all the line segments of the building boundaries are either parallel or perpendicular. However, this assumption is not valid for many buildings. We present a new approach consisting of three steps that is applicable to more complex building boundaries. The three steps consist of boundary tracing, generalization, and regularization. Each step contains algorithms that range from slight modifications of conventional algorithms to entirely new concepts. Four typical building shapes were selected to test the performance of out new approach and the results were compared with digital maps. The results show that the proposed approach has good potential for extracting building boundaries of various shapes.