• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.10 no.3
• /
• pp.419-429
• /
• 1999

Identifying the boundary of the effective receiving power of waves is one of the most important factors for cell optimization. In this paper, we introduce a propagation loss prediction model which yields highly accurate prediction in very complex areas as Seoul where a mixture of many large buildings, small buildings, broad streets, narrow alleys, rivers and forests co-exist in an irregular arrangement. This prediction model is based on neural networks trained on field measurement data collected in the past. Using these data along with 3-D digital elevation maps and vector data for building structures, we extract the parameter values which mainly affect the amount of propagation loss. These parameter values are then used as the inputs to the neural network. Trained neural network becomes the approximated function of the propagation loss model which generalizes very well and can predict accurately in the regions not included in training the neural network. The experimental results show a superior performance over the other models in the cells operating in the city of Seoul.

• Journal of KIISE:Information Networking
• /
• v.30 no.3
• /
• pp.343-354
• /
• 2003

This paper describes algorithm that minimizes recovery time of packet loss resulting from handoff in multicast environments and guarantees reliability through interaction of FN(Foreign Network) with PMTP(Predictable Multicast Tree Protocol). To solve the problems that inefficient routing and handoff delay taking plate when using hi-directional tunneling and remote subscription independently in multicast environments, proposed algorithm uses tunneling and rejoining multicast group according to the status of an arriving FA in a foreign network. Furthermore, proposed algorithm sends packet loss information and register message to previous FA or current FA at the same time. so, MH is able to recovery packet loss in handoff delay as soon as possible. As a result of performance analysis, proposed algorithm is more efficient than previous researches and is applicable to existing handoff method without requiring additional procedures.

• Journal of KIISE:Information Networking
• /
• v.29 no.4
• /
• pp.437-446
• /
• 2002

In order to reduce the increasing packet loss rates caused by an exponential increase in network traffic, the IETF (Internet Engineering Task Force) is considering the deployment of active queue management techniques such as RED (Random Early Detection). While active queue management in routers and gateways can potentially reduce packet loss rates in the Internet, this paper has demonstrated the inherent weakness of current techniques and shows that they are ineffective in preventing high loss rates. The inherent problem with these queue management algorithms is that they all use static parameter setting. So, in case where these parameters do not match the requirement of the network load, the performance of these algorithms can approach that of a traditional Drop-tail. In this paper, in order to solve this problem, a new active queue management algorithm called ARED (Active RED) is proposed. ARED computes the parameter based on our heuristic method. This algorithm can effectively reduce packet loss while maintaining high link utilizations.

• Journal of Korea Multimedia Society
• /
• v.25 no.10
• /
• pp.1375-1385
• /
• 2022

In this paper, we propose a semi-supervised domain adaptation solution to deal with practical face recognition (FR) scenarios where a single face image for each target identity (to be recognized) is only available in the training phase. Main goal of the proposed method is to reduce the discrepancy between the target and the source domain face images, which ultimately improves FR performances. The proposed method is based on the Domain Adatation network (DAN) using an MMD loss function to reduce the discrepancy between domains. In order to train more effectively, we develop a novel loss function learning strategy in which MMD loss and cross-entropy loss functions are adopted by using different weights according to the progress of each epoch during the learning. The proposed weight adoptation focuses on the training of the source domain in the initial learning phase to learn facial feature information such as eyes, nose, and mouth. After the initial learning is completed, the resulting feature information is used to training a deep network using the target domain images. To evaluate the effectiveness of the proposed method, FR performances were evaluated with pretrained model trained only with CASIA-webface (source images) and fine-tuned model trained only with FERET's gallery (target images) under the same FR scenarios. The experimental results showed that the proposed semi-supervised domain adaptation can be improved by 24.78% compared to the pre-trained model and 28.42% compared to the fine-tuned model. In addition, the proposed method outperformed other state-of-the-arts domain adaptation approaches by 9.41%.

• Smart Structures and Systems
• /
• v.30 no.6
• /
• pp.673-686
• /
• 2022

Deep learning algorithms for Structural Health Monitoring (SHM) have been extracting the interest of researchers and engineers. These algorithms commonly used loss functions and evaluation indices like the mean square error (MSE) which were not originally designed for SHM problems. An updated loss function which was specifically constructed for deep-learning-based structural damage detection problems has been proposed in this study. By tuning the coefficients of the loss function, the weights for damage localization and quantification can be adapted to the real situation and the deep learning network can avoid unnecessary iterations on damage localization and focus on the damage severity identification. To prove efficiency of the proposed method, structural damage detection using convolutional neural networks (CNNs) was conducted on a truss bridge model. Results showed that the validation curve with the updated loss function converged faster than the traditional MSE. Data augmentation was conducted to improve the anti-noise ability of the proposed method. For reducing the training time, the normalized modal strain energy change (NMSEC) was extracted, and the principal component analysis (PCA) was adopted for dimension reduction. The results showed that the training time was reduced by 90% and the damage identification accuracy could also have a slight increase. Furthermore, the effect of different modes and elements on the training dataset was also analyzed. The proposed method could greatly improve the performance for structural damage detection on both the training time and detection accuracy.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.17 no.5
• /
• pp.119-124
• /
• 2017

In 1997, IBM's DeepBlue won the world chess championship, Garry Kasparov, and recently, Google's AlphaGo won all three games against Ke Jie, who was ranked 1st among all human Baduk players worldwide, interest in deep running has increased rapidly. DeepPurple, proposed in this paper, is a AI chess engine based on deep learning. DeepPurple Chess Engine consists largely of Monte Carlo Tree Search and policy network and value network, which are implemented by convolution neural networks. Through the policy network, the next move is predicted and the given situation is calculated through the value network. To select the most beneficial next move Monte Carlo Tree Search is used. The results show that the accuracy and the loss function cost of the policy network is 43% and 1.9. In the case of the value network, the accuracy is 50% and the loss function cost is 1, respectively.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.7
• /
• pp.1706-1725
• /
• 2024

The unique U-shaped structure of U-Net network makes it achieve good performance in image segmentation. This network is a lightweight network with a small number of parameters for small image segmentation datasets. However, when the medical image to be segmented contains a lot of detailed information, the segmentation results cannot fully meet the actual requirements. In order to achieve higher accuracy of medical image segmentation, a novel improved U-Net network architecture called multi-scale encoder-decoder U-Net+ (MEDU-Net+) is proposed in this paper. We design the GoogLeNet for achieving more information at the encoder of the proposed MEDU-Net+, and present the multi-scale feature extraction for fusing semantic information of different scales in the encoder and decoder. Meanwhile, we also introduce the layer-by-layer skip connection to connect the information of each layer, so that there is no need to encode the last layer and return the information. The proposed MEDU-Net+ divides the unknown depth network into each part of deconvolution layer to replace the direct connection of the encoder and decoder in U-Net. In addition, a new combined loss function is proposed to extract more edge information by combining the advantages of the generalized dice and the focal loss functions. Finally, we validate our proposed MEDU-Net+ MEDU-Net+ and other classic medical image segmentation networks on three medical image datasets. The experimental results show that our proposed MEDU-Net+ has prominent superior performance compared with other medical image segmentation networks.

• 한국정보통신설비학회:학술대회논문집
• /
• 2007.08a
• /
• pp.327-330
• /
• 2007

This article discusses a capacity planning method in QoS-guaranteed IP networks such as BcN (Broadband convergence Network). Since IP based networks have been developed to transport best-effort data traffic, the introduction of multi-service component in BcN requires fundamental modifications in capacity planning and network dimensioning. In this article, we present the key issues of the capacity planning in multi-service IP networks. To provide a foundation for network dimensioning procedure, we describe a systematic approach for classification and modeling of BcN traffic based on the QoS requirements of BcN services. We propose a capacity planning framework considering data traffic and real-time streaming traffic separately. The multi-service Erlang model, an extension of the conventional Erlang B loss model, is introduced to determine required link capacity for the call based real-time streaming traffic. The application of multi-service Erlang model can provide significant improvement in network planning due to sharing of network bandwidth among the different services.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 1996.10a
• /
• pp.186-189
• /
• 1996

In most of the recent research literature, network performance is expressed in terms of network engineering measures such as delay or loss. These performance measures are important to network owners and operators, but it is believed that user preferences should be the primary consideration which drives the resource allocation scheme. A network is only as valuable as its users perceive it to be. Therefore, it is advocated that the users themselves determine relative traffic priorities. This paper describes the role of feedback in network resource allocation, which could be part of a user-oriented framework for network operation and control. Feedback mechanism can also be used to improve the two types of efficiency in the network; network efficiency and economic efficiency.

• Proceedings of the Korean Information Science Society Conference
• /
• 2002.10e
• /
• pp.76-78
• /
• 2002

Multicasting over mobile IP network becomes more important with the increasing needs of supporting multimedia services in mobile network. The IETF has suggested two approaches which are remote subscription and bidirectional tunneling for supporting mobility management in multicasting over mobile IP. But these protocols have problems - the frequent reconstruction of multicast delivery tree, packet less during handoff, convergence problem, and so on. In this paper, we propose to use preconfiguration of multicast delivery tree when mobile host enters the foreign network. It will decrease the frequency of multicast delivery tree reconstruction, and reduce the packet loss during handoff, Also the multicast delivery tree maintained by Keep Alive messages makes the signaling overload of networks diminished.