• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.7 no.5
• /
• pp.1252-1271
• /
• 2013

Automatic image annotation has become an increasingly important research topic owing to its key role in image retrieval. Simultaneously, it is highly challenging when facing to large-scale dataset with large variance. Practical approaches generally rely on similarity measures defined over images and multi-label prediction methods. More specifically, those approaches usually 1) leverage similarity measures predefined or learned by optimizing for ranking or annotation, which might be not adaptive enough to datasets; and 2) predict labels separately without taking the correlation of labels into account. In this paper, we propose a method for image annotation through collaborative similarity metric learning from dataset and modeling the label correlation of the dataset. The similarity metric is learned by simultaneously optimizing the 1) image ranking using structural SVM (SSVM), and 2) image annotation using correlated label propagation, with respect to the similarity metric. The learned similarity metric, fully exploiting the available information of datasets, would improve the two collaborative components, ranking and annotation, and sequentially the retrieval system itself. We evaluated the proposed method on Corel5k, Corel30k and EspGame databases. The results for annotation and retrieval show the competitive performance of the proposed method.

• Journal of KIISE:Information Networking
• /
• v.30 no.6
• /
• pp.787-798
• /
• 2003

The flow label field in IPv6 has been proposed to provide the QoS. Since the existing flow label specification scheme like random-number format utilizes the label only as the identifier of flow, it is not appropriate for providing differentiated services according to the characteristics of various types of real-time traffic. This paper proposes a hybrid scheme that makes use of the flow label fields as components of flow and QoS parameters as well. To be specific, this paper investigates a scheme that both guarantees the end-to-end service quality and utilizes efficiently backbone resources by allowing users to specify QoS parameters using flow labels. Assuming an MLPS-TE network as the backbone, we compare the performance of our proposed scheme with that of random-number scheme through simulation. The simulation result shows that our scheme is more efficient than the existing one in terms of the transmission rate as well as the resource utilization of the backbone.

• Journal of the Korean Society of Food Culture
• /
• v.25 no.6
• /
• pp.725-733
• /
• 2010

This study examined the understanding and use of information on nutrition labels based on one serving size among female consumers above the age of 20 in Seoul area. According to the survey, 69.9% of respondents were aware of the current system of nutrition labeling based on one serving size, and 51.8% of the respondents expressed their dissatisfaction with the system because the nutrition labels were difficult to understand or appeared unreliable. The nutrition label literacy of the consumers varied with respect to different packaging units. The respondents were likely to be less accurate in calculating the expected caloric intake when only portions of a multi-serving package were used. Initially 69.0% of respondents reported that they had read the nutrition label before purchasing a product but 91.9% of respondents said that they would check the label after learning how to read the label properly. It is very important to make consumers aware that the labels are very reliable sources of nutrition information. A public education campaign on the use of nutrition labels should focus on developing the consumers' ability and skills in using the label information when choosing foods.

• 한국어정보학회:학술대회논문집
• /
• 2017.10a
• /
• pp.56-59
• /
• 2017

CNN(Convolutional Neural Network)을 이용하여 발화 주제 다중 분류 task를 multi-labeling 방법과, cluster 방법을 이용하여 수행하고, 각 방법론에 MSE(Mean Square Error), softmax cross-entropy, sigmoid cross-entropy를 적용하여 성능을 평가하였다. Network는 음절 단위로 tokenize하고, 품사정보를 각 token의 추가한 sequence와, Naver DB를 통하여 얻은 named entity 정보를 입력으로 사용한다. 실험결과 cluster 방법으로 문제를 변형하고, sigmoid를 output layer의 activation function으로 사용하고 cross entropy cost function을 이용하여 network를 학습시켰을 때 F1 0.9873으로 가장 좋은 성능을 보였다.

• Journal of Internet Computing and Services
• /
• v.18 no.1
• /
• pp.77-88
• /
• 2017

Recently, with the advent of knowledge based society where information and knowledge make values, patents which are the representative form of intellectual property have become important, and the number of the patents follows growing trends. Thus, it needs to classify the patents depending on the technological topic of the invention appropriately in order to use a vast amount of the patent information effectively. IPC (International Patent Classification) is widely used for this situation. Researches about IPC automatic classification have been studied using data mining and machine learning algorithms to improve current IPC classification task which categorizes patent documents by hand. However, most of the previous researches have focused on applying various existing machine learning methods to the patent documents rather than considering on the characteristics of the data or the structure of patent documents. In this paper, therefore, we propose to use two structural fields, technical field and background, considered as having impacts on the patent classification, where the two field are selected by applying of the characteristics of patent documents and the role of the structural fields. We also construct multi-label classification model to reflect what a patent document could have multiple IPCs. Furthermore, we propose a method to classify patent documents at the IPC subclass level comprised of 630 categories so that we investigate the possibility of applying the IPC multi-label classification model into the real field. The effect of structural fields of patent documents are examined using 564,793 registered patents in Korea, and 87.2% precision is obtained in the case of using title, abstract, claims, technical field and background. From this sequence, we verify that the technical field and background have an important role in improving the precision of IPC multi-label classification in IPC subclass level.

• Convergence Security Journal
• /
• v.23 no.1
• /
• pp.3-8
• /
• 2023

Multi-domain optical transport networks are not fundamentally interoperable and require an integrated orchestration mechanism and path provision mechanism at the entire network level. In addition, ensuring network survivability is one of the important issues. MPLS-TP (Multi-Protocol Label Switching-Transport Profile) defines various protection/recovery methods as standards, but does not mention how to calculate and select protection/recovery paths. Therefore, an algorithm that minimizes protection/recovery collisions at the optical circuit packet integrated network level and calculates and sets a path that can be rapidly protected/recovered over the entire integrated network area is required. In this paper, we proposed an algorithm that calculates and sets up a path that can be rapidly protected and restored in a T-SDN network composed of multiple ring-mesh topology.

• Journal of the Institute of Electronics Engineers of Korea TC
• /
• v.41 no.6 s.324
• /
• pp.69-78
• /
• 2004

This paper proposes architecture and characteristics of a multi-ring based optical network with single-hop between edge nodes using the concept of circuit switching and multi-wavelength label switching to solve delay problem caused by applying crossconnectors as transit nodes in the wavelength division multiplexing(WDM) network. We suggest multi-ring based architecture composed single and multiple wavelength-bands with multi-wavelength labels, and analyze characteristics of two models. To avoid the packet collision in output ports of edge nodes due to output contention, the static and dynamic allocation scheme, which packets are allocated in time slots, is provided. Based on our analysis, it shows that delay only occur in not core nodes but edge nodes in the proposed architecture. In addition, we evaluate the probabilities of delay, packet loss, and call blocking in the proposed optical packet network.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.35 no.8B
• /
• pp.1141-1149
• /
• 2010

Next Generation Networks (NGN) is defined as IP-based networks with multi-services and with multi-access networks. A variety of services and access technologies are co-existed within NGN. Therefore there are numerous transport technologies such as Differentiated Services (DiffServ), Multi-protocol Label Switching (MPLS), and the combined transport technologies. In such an environment, flows are aggregated and de-aggregated multiple times in their end-to-end paths. In this research, a method for calculating end-to-end delay bound for such a flow, provided that the information exchanged among networks regarding flow aggregates, especially the maximum burst size of a flow aggregate entering a network. We suggest an admission control mechanism that can decide whether the requested performance for a flow can be met. We further verify the suggested calculation and admission algorithm with a few realistic scenarios.

• The Journal of the Korea Contents Association
• /
• v.20 no.1
• /
• pp.324-333
• /
• 2020

This study compares the visual elements of the top five global beer brands currently in the Chinese market with those of the top five Chinese beer brands. The purpose of the research is to identify the importance of visual elements in the label design of beer package and to present the direction of efficient label design. To this end, it is a brand among visual elements of beer label design, and it investigates the label design of Chinese beer and global beer brand through the multi-accelerated comparative analysis method, focusing on color, illustration and layout, and analyzes the difference of elements appearing in the global beer label design from China by utilizing the recursive five-point scale. According to the research, the visual elements used in the Chinese beer brand were more complex compared to the global beer brand, and most layouts were similar, making the product less differentiated. On the other hand, the global beer brand has secured brand identity and differentiation based on its iconic visual elements and diversity of layout. In order to establish itself as a strong brand in consumers' minds, the Chinese beer brand needs differentiated label design through active use of unique brand design.

• Information and Communications Magazine
• /
• v.18 no.11
• /
• pp.44-57
• /
• 2001

본 논문에서는 광전달망에서 토의가 진행되고 있는 경로 라우팅 문제와 시그널링 프로토콜 기술에 관하여 기술한다. 광전달망의 구조는 백본 네트워크 코어부분에 광회선분배기(Optical Crossconnect; OXC)가 있고 에지의 IP 라우터와 연결되는 형태로 가정한다. Generalized Multi-Protocol Label Stwitch(GMPLS) 기술을 근간으로 하고 있다 특히 본 논문에서는 IP 라우팅과 시그널링을 광전달망에 적용하기 위해 필요한 기술들에 대하여 기술한다.