• Journal of Korea Multimedia Society
• /
• v.7 no.7
• /
• pp.905-912
• /
• 2004

It is a study LVMPD to solve the fairness problem of inter-session shared bandwidth. Whenever congestion occurs in one session, the highest layer is dropped. Also the highest layer of other sessions in non-congestion is dropped by iner-session fairness. While there is spare bandwidth, receivers of session in non-congestion can't use bandwidth efficiently. In this paper, we proposed a distribution method to use efficiently spare bandwidth that occurred by inter-session fairness. In our method, we considered the highest layer those receivers receiving and the higher layer those receivers requesting. The sender assigns the packet-deleting priority to packets when congestion occurs in receivers, and sets layer add/drop flag when receivers of session in non-congestion request the higher layer. The switch deletes packets with packet-deleting priority and transfers packets with layer add/drop flag for session in non-congestion. Therefore receivers of session in non-congestion can add the higher layer. In experimental results, it was known that proposed algorithm use the inter-session shared bandwidth more effectively compared with already known method.

• Journal of KIISE:Computing Practices and Letters
• /
• v.14 no.2
• /
• pp.181-185
• /
• 2008

Layered video multicast such as RLM Receiver-driven layered multicast) is a premising technique for delivering streaming video to a set of heterogeneous receivers over ALM(Application Layer Multicast) as well as over IP multicast. However, this approach may suffer from unnecessary fluctuation of video quality due to overlapped and failed join-experiments. Though a shared teaming scheme was introduced to resolve these problems, it may cause high control overhead and slow convergence problem when used with ALM. In this paper, we propose a new shared learning scheme for ALM-based layered video multicast which reduces control overhead and convergence latency while keeping the number of fluctuation reasonably small. The simulation results show that the proposed scheme performs better than an ALM-based layered video multicast with shared learning in terms of control overhead and convergence latency.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2017.05a
• /
• pp.645-648
• /
• 2017

The existing embedded tester used TCP and shared file system for log processing. In addition, the existing processing method was treated as 1-N structure. This method wastes resources of the tester for exception handling. We implemented a log processing message layer that can be distributed by messaging system. And we compare the transmission method using the message layer and the transmission method using TCP and the shared file system. As a result of comparison, transmission using the message layer showed higher transmission bandwidth than TCP. In the CPU usage, the message layer showed lower efficiency than TCP, but showed no significant difference. It can be seen that the log processing using the message layer shows higher efficiency.

• Journal of Information Processing Systems
• /
• v.19 no.1
• /
• pp.130-138
• /
• 2023

The traffic flow in an urban area is affected by the date, weather, and regional traffic flow. The existing methods are weak to model the dynamic road network features, which results in inadequate long-term prediction performance. To solve the problems regarding insufficient capacity for dynamic modeling of road network structures and insufficient mining of dynamic spatio-temporal features. In this study, we propose a novel traffic flow prediction framework called shared spatio-temporal attention convolution optimization network (SSTACON). The shared spatio-temporal attention convolution layer shares a spatio-temporal attention structure, that is designed to extract dynamic spatio-temporal features from historical traffic conditions. Subsequently, the graph optimization module is used to model the dynamic road network structure. The experimental evaluation conducted on two datasets shows that the proposed method outperforms state-of-the-art methods at all time intervals.

• Journal of KIISE
• /
• v.44 no.12
• /
• pp.1252-1257
• /
• 2017

In dialogues, speakers' intentions can be represented by sets of an emotion, a speech act, and a predicator. Therefore, dialogue systems should capture and process these implied characteristics of utterances. Many previous studies have considered such determination as independent classification problems, but others have showed them to be associated with each other. In this paper, we propose an integrated model that simultaneously determines emotions, speech acts, and predicators using a convolution neural network. The proposed model consists of a particular abstraction layer, mutually independent informations of these characteristics are abstracted. In the shared abstraction layer, combinations of the independent information is abstracted. During training, errors of emotions, errors of speech acts, and errors of predicators are partially back-propagated through the layers. In the experiments, the proposed integrated model showed better performances (2%p in emotion determination, 11%p in speech act determination, and 3%p in predicator determination) than independent determination models.

• Proceedings of the Korean Information Science Society Conference
• /
• 2007.06a
• /
• pp.196-197
• /
• 2007

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.6 no.1
• /
• pp.178-202
• /
• 2012

In this paper, we first survey cross-layer architectures for Wireless Sensor Networks (WSNs) and Wireless Multimedia Sensor Networks (WMSNs). Afterwards, we propose a novel cross-layer architecture for QoS provisioning in clustered and multi-hop based WMSNs. The proposed architecture provides support for multiple network-based applications on a single sensor node. For supporting multiple applications on a single node, an area in memory is reserved where each application can store its network protocols settings. Furthermore, the proposed cross-layer architecture supports heterogeneous flows by classifying WMSN traffic into six traffic classes. The architecture incorporates a service differentiation module for QoS provisioning in WMSNs. The service differentiation module defines the forwarding behavior corresponding to each traffic class. The forwarding behavior is primarily determined by the priority of the traffic class, moreover the service differentiation module allocates bandwidth to each traffic class with goals to maximize network utilization and avoid starvation of low priority flows. The proposal incorporates the congestion detection and control algorithm. Upon detection of congestion, the congested node makes an estimate of the data rate that should be used by the node itself and its one-hop away upstream nodes. While estimating the data rate, the congested node considers the characteristics of different traffic classes along with their total bandwidth usage. The architecture uses a shared database to enable cross-layer interactions. Application's network protocol settings and the interaction with the shared database is done through a cross-layer optimization middleware.

• ETRI Journal
• /
• v.27 no.6
• /
• pp.725-732
• /
• 2005

The reconfiguration management scheme changes a logical topology in response to changing traffic patterns in the higher layer of a network or the congestion level on the logical topology. In this paper, we formulate a reconfiguration scheme with a shared buffer-constrained cost model based on required quality-of-service (QoS) constraints, reconfiguration penalty cost, and buffer gain cost through traffic aggregation. The proposed scheme maximizes the derived expected reward-cost function as well as guarantees the required flow's QoS. Simulation results show that our reconfiguration scheme significantly outperforms the conventional one, while the required physical resources are limited.

• Journal of the Korea Society of Computer and Information
• /
• v.21 no.1
• /
• pp.1-8
• /
• 2016

Convolutional neural network (CNN) consists of a few pairs of both convolution layer and subsampling layer. Thus it has more hidden layers than multi-layer perceptron. With the increased layers, the size of convolution mask ultimately determines the total number of weights in CNN because the mask is shared among input images. It also is an important learning factor which makes or breaks CNN's learning. Therefore, this paper proposes the best method to choose the convolution size and the number of layers for learning CNN successfully. Through our face recognition with vast learning examples, we found that the best size of convolution mask is 5 by 5 and 7 by 7, regardless of the number of layers. In addition, the CNN with two pairs of both convolution and subsampling layer is found to make the best performance as if the multi-layer perceptron having two hidden layers does.

• Journal of KIISE:Computer Systems and Theory
• /
• v.30 no.7_8
• /
• pp.331-340
• /
• 2003

Low-cost commodity SMP(Symmetric Multiprocessor) is widely used as a node of cluster system. In this paper, we implement and evaluate the performance of SDSM system for SMP clusters. Our SDSM system provides HLRC(Home-based Lazy Release Consistency) memory consistency model. Our protocol utilize shared memory within same SMP node, so that page fetch and message passing through network can be reduced. It is implemented on 8 node of 2-way Pentium-III SMP interconnected with 100Mbps Fast Ethernet, and uses TCP/IP for transport/network layer protocol. The experiment with eight applications shows that our SMP protocol achieves maximum 33% speedup improvement and 13%-52% reduction of page fetch compared with uniprocessor protocol.