• Journal of The Korean Association For Science Education
• /
• v.34 no.3
• /
• pp.295-301
• /
• 2014

The purpose of this study is to investigate the difficulties of teachers and students on the unit about 'measuring weight.' In this research, we have acquired data about teachers through survey, interview, and self-reflection journals, at the same time we have collected information on the students through survey, assessment test, and interview. We have extracted the difficulties from analysis with constant comparison method. In addition, we have analysed the curriculum of science and mathematics to know the leaning sequence. The analysis had been checked up by experts in science education. The result of the study is as follows: The difficulties of teachers are from the lack of teachers' descriptive knowledge, disorder of conceptual hierarchy in the curriculum, poor experimental instruments, and low psychomotor skill of students. The difficulties of students are from common misconceptions, opaque concepts, lack of manipulation skill, insufficiency of mathematical ability, difficulty of application of principles to the real situation, and lack of problem-solving ability. In addition, teachers have recognized that students face more difficulties in experiment class, while students think that they face more difficulties in conceptual understanding class.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.24 no.2
• /
• pp.161-165
• /
• 2014

In this paper, we proposed the fuzzy prototype pattern classifier. In the proposed classifier, each prototype is defined to describe the related sub-space and the weight value is assigned to the prototype. The weight value assigned to the prototype leads to the change of the boundary surface. In order to define the prototypes, we use Fuzzy C-Means Clustering which is the one of fuzzy clustering methods. In order to optimize the weight values assigned to the prototypes, we use the Differential Evolutionary Algorithm. We use Linear Discriminant Analysis to estimate the coefficients of the polynomial which is the structure of the consequent part of a fuzzy rule. Finally, in order to evaluate the classification ability of the proposed pattern classifier, the machine learning data sets are used.

• Journal of the Korean Data and Information Science Society
• /
• v.27 no.5
• /
• pp.1273-1284
• /
• 2016

In this paper, we analyzed the patents on machine learning using keyword network analysis and clique analysis. To construct a keyword network, important keywords were extracted based on the TF-IDF weight and their association, and network structure analysis and clique analysis was performed. Density and clustering coefficient of the patent keyword network are low, which shows that patent keywords on machine learning are weakly connected with each other. It is because the important patents on machine learning are mainly registered in the application system of machine learning rather thant machine learning techniques. Also, our results of clique analysis showed that the keywords found by cliques in 2005 patents are the subjects such as newsmaker verification, product forecasting, virus detection, biomarkers, and workflow management, while those in 2015 patents contain the subjects such as digital imaging, payment card, calling system, mammogram system, price prediction, etc. The clique analysis can be used not only for identifying specialized subjects, but also for search keywords in patent search systems.

• (The)Study of the Eastern Classic
• /
• no.36
• /
• pp.265-299
• /
• 2009

The Great Learning is an essential scripture of Confucianism that has had great influence on the politics, thoughts, society and culture of the East Asia which contains Korea, Japan and China. In case of Japan, the doctrine of Toegye(退溪) is flown into pre-Tokugawa era, and as the doctrine of Zhu Xi(朱熹) exercise influences over the other thoughts, the importance of the Great Learning is embossed relatively in Tokugawa era. The characteristic of Japanese confucianism of Tokugawa era is to lay weight on real world as such, and the Japanese confucianism has grown up academically centers on exhaustive reinterpretation about some Confucian classics. And Backho-Yoon Hyu(白湖 尹?), Seokye-Park Sedang(西溪 朴世堂) who lived in 17th century of Joseon dynasty attempt new interpretation about the Great Learning and they have an objection to the explanatory notes of Zhu Xi. In the same period of China, there are similar academic trends around Whang Jong Hi(黃宗羲), Ko Yeom Mu(顧炎武), Wang Fu Chi(王夫之). In other words, new views of real scholarship which reject emptiness and put much value on reality were current of thoughts that have been common to the oriental three countries in early and middle 17th century. The main object of this paper is to understand the attitude and understanding about the Confucian classics especially the Great Learning of the scholars who lived in early and middle Tokugawa era. It will be a decisive clue to understand the ninucture of thoughts system of them. Through these work, we understand how it has had influenced to thinking-ninucture and lives of the Japanese. And the other purpose of this paper is to understand characteristics of them when we compare that vking-ninucture and lives of the Japanese Confuciang-ninearly and middle Tokugawa era with in the same period of Jeseon and the later Ming(明) and early Ching(淸) dynasty.

• Journal of Intelligence and Information Systems
• /
• v.28 no.2
• /
• pp.101-125
• /
• 2022

Recently, with the development of computing technology and the improvement of the cloud environment, deep learning technology has developed, and attempts to apply deep learning to various fields are increasing. A typical example is anomaly detection, which is a technique for identifying values or patterns that deviate from normal data. Among the representative types of anomaly detection, it is very difficult to detect a contextual anomaly that requires understanding of the overall situation. In general, detection of anomalies in image data is performed using a pre-trained model trained on large data. However, since this pre-trained model was created by focusing on object classification of images, there is a limit to be applied to anomaly detection that needs to understand complex situations created by various objects. Therefore, in this study, we newly propose a two-step pre-trained model for detecting abnormal situation. Our methodology performs additional learning from image captioning to understand not only mere objects but also the complicated situation created by them. Specifically, the proposed methodology transfers knowledge of the pre-trained model that has learned object classification with ImageNet data to the image captioning model, and uses the caption that describes the situation represented by the image. Afterwards, the weight obtained by learning the situational characteristics through images and captions is extracted and fine-tuning is performed to generate an anomaly detection model. To evaluate the performance of the proposed methodology, an anomaly detection experiment was performed on 400 situational images and the experimental results showed that the proposed methodology was superior in terms of anomaly detection accuracy and F1-score compared to the existing traditional pre-trained model.

• Journal of the Korean Society of Radiology
• /
• v.18 no.1
• /
• pp.37-44
• /
• 2024

Even though MR can reveal excellent soft-tissue contrast and functional information, CT is also required for electron density information for accurate dose calculation in Radiotherapy. For the fusion of MRI and CT images in RT treatment planning workflow, patients are normally scanned on both MRI and CT imaging modalities. Recently deep-learning-based generations of CT images from MR images became possible owing to machine learning technology. This eliminated CT scanning work. This study implemented a CycleGan deep-learning-based CT image generation from MR images. Three CT generators whose learning is based on T1- , T2- , or T1-&T2-weighted MR images were created, respectively. We found that the T1-weighted MR image-based generator can generate better than other CT generators when T1-weighted MR images are input. In contrast, a T2-weighted MR image-based generator can generate better than other CT generators do when T2-weighted MR images are input. The results say that the CT generator from MR images is just outside the practical clinics and the specific weight MR image-based machine-learning generator can generate better CT images than other sequence MR image-based generators do.

• Journal of the Korea Society of Computer and Information
• /
• v.25 no.3
• /
• pp.19-26
• /
• 2020

Steel production requires high-quality surfaces with minimal defects. Therefore, the detection algorithms for the surface defects of steel strip should have good generalization performance. To meet the growing demand for high-quality products, the use of intelligent visual inspection systems is becoming essential in production lines. In this paper, we proposed a ShuffleDefectNet defect detection system based on deep learning. The proposed defect detection system exceeds state-of-the-art performance for defect detection on the Northeastern University (NEU) dataset obtaining a mean average accuracy of 99.75%. We train the best performing detection with different amounts of training data and observe the performance of detection. We notice that accuracy and speed improve significantly when use the overall architecture of ShuffleDefectNet.

• Proceedings of the KIEE Conference
• /
• 2001.11c
• /
• pp.431-434
• /
• 2001

단층 퍼셉트론이 처음 개발되었을 때, 간단한 패턴을 인식하는 학습 기능을 가지고 있기 장점 때문에 학자들의 관심을 끌었다. 단층 퍼셉트론은 한 개의 소자를 이용해서 이진 논리를 가중치(weight)의 변경만으로 모두 표현할 수 있는 장점 때문에 영상처리, 패턴인식, 장면인식 등에 이용되어 왔다. 최근에, 역전파학습(Back-Propagation Learning)알고리즘이 이진 공간내의 매핑 문제에 적용되고 있다. 그러나, 역전파 학습알고리즘은 연속공간 내에서 긴 학습시간과 비효율적인 수행의 문제를 가지고 있다. 일반적으로 역전파 학습 알고리즘은 간단한 이진 공간에서 매핑하기 위해서 많은 반복과정을 요구한다. 역전파 학습 알고리즘에서는 은닉층의 뉴런의 수는 주어진 문제를 해결하기 위해서 우선순위(prior)를 알지 못하기 때문에 입력층과 출력층내의 뉴런의 수에 의존한다. 따라서, 3층 신경회로망의 적용에 있어 가장 중요한 문제중의 하나는 은닉층내의 필요한 뉴런수를 결정하는 것이고, 회로망 합성과 가중치 결정에 대한 적절한 방법을 찾지 못해 실제로 그 사용 영역이 한정되어 있었다. 본 논문에서는 패턴 분류를 위한 새로운 학습방법을 제시한다. 훈련입력의 기하학적인 분석에 기반을 둔 이진 신경회로망내의 은닉층내의 뉴런의 수를 자동적으로 결정할 수 있는 NETLA(Newly Expand and Truncate Learning Algorithm)라 불리우는 기하학적 학습알고리즘을 제시하고, 시뮬레이션을 통하여, 제안한 알고리즘의 우수성을 증명한다.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2017.01a
• /
• pp.151-154
• /
• 2017

소프트웨어의 개발과정 중 결함을 제거하는 작업인 디버깅을 위해서는 가장 먼저 그 결함의 정확한 위치를 찾아야한다. 이 작업은 많은 시간이 소요되며, 이 시간을 단축시키기 위한 결함 위치 식별 기법들이 소개되었다. 많은 기법들 중 프로그램 커버리지 정보를 학습하여 규칙을 분석하는 인공신경망 기반 선행 연구가 있다. 이를 기반으로 본 논문에서는 문장들 간의 관계를 추가적으로 파악하여 학습 데이터로 사용하는 기법을 제안한다. 특정 문장이 항상 지나는 테스트케이스들 중 나머지 다른 문장들이 지나는 테스트케이스의 비율을 통해 문장들 간의 관계를 나타낸다. 해당 비율을 계산하기 위해 조건부 확률인 베이지안 확률을 사용한다. 베이지안 확률을 통해 얻은 문장들의 관계에 따라 인공신경망 내에서 의심도를 결정하는 웨이트(weight)가 기존 기법과는 다르게 학습된다. 이 차이는 문장들의 의심도를 조정하며, 결과적으로 다중 결함 위치 식별의 정확도를 향상시킨다. 본 논문에서 제안한 기법을 이용하여 실험한 결과, Tarantula 대비 평균 39.8%, 기존 역전파 인공신경망(BPNN) 기반 기법 대비 평균 60.5%의 정확도 향상이 있었음을 확인할 수 있다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.3
• /
• pp.976-990
• /
• 2020

The k-means algorithm is widely used in academia and industry due to easy and simple implementation, enabling fast learning for complex datasets. However, k-means struggles to classify datasets without prior knowledge of specific domains. We proposed the repulsive k-means (RK-means) algorithm in a previous study to improve the k-means algorithm, using the repulsive force concept, which allows deleting unnecessary cluster centroids. Accordingly, the RK-means enables to classifying of a dataset without domain knowledge. However, three main problems remain. The RK-means algorithm includes a cluster repulsive force offset, for clusters confined in other clusters, which can cause cluster locking; we were unable to prove RK-means provided optimal convergence in the previous study; and RK-means shown better performance only normalize term and weight. Therefore, this paper proposes the advanced RK-means (ARK-means) algorithm to resolve the RK-means problems. We establish an initialization strategy for deploying cluster centroids and define a metric for the ARK-means algorithm. Finally, we redefine the mass and normalize terms to close to the general dataset. We show ARK-means feasibility experimentally using blob and iris datasets. Experiment results verify the proposed ARK-means algorithm provides better performance than k-means, k'-means, and RK-means.