• Journal of Korea Society of Industrial Information Systems
• /
• v.22 no.5
• /
• pp.17-24
• /
• 2017

Recently Deep-learning, Technologies using Large or Deep Artificial Neural Networks, have Shown Remarkable Performance, and the Increasing Size of the Network Contributes to its Performance Improvement. However, the Increase in the Size of the Neural Network Leads to an Increase in the Calculation Amount, which Causes Problems Such as Circuit Complexity, Price, Heat Generation, and Real-time Restriction. In This Paper, We Propose and Test a Method to Reduce the Number of Network Connections by Effectively Pruning the Redundancy in the Connection and Showing the Difference between the Performance and the Desired Range of the Original Neural Network. In Particular, we Proposed a Simple Method to Improve the Performance by Re-learning and to Guarantee the Desired Performance by Allocating the Error Rate per Layer in Order to Consider the Difference of each Layer. Experiments have been Performed on a Typical Neural Network Structure such as FCN (full connection network) and CNN (convolution neural network) Structure and Confirmed that the Performance Similar to that of the Original Neural Network can be Obtained by Only about 1/10 Connection.

• Journal of KIISE:Information Networking
• /
• v.31 no.6
• /
• pp.659-672
• /
• 2004

A Mobile Ad hoc Network (MANET) is a group of independent mobile computing nodes that consist of a multi-hop wireless network without a central administration or any infrastructure. Every node that wants to join a MANET must obtain an address for communication. Having a centralized DHCP server that provides addresses to nodes, we can easily and automatically obtain addresses. However, a MANET lacks any fixed infrastructure such as a DHCP server. We therefore propose a distributed address autoconfiguration approach for a MANET using a reserved address and optimistic Duplicated Address Detection (DAD). The reserved address helps to reduce the allocation latency, and the optimistic DAD guarantees the uniqueness of addresses and lessens communication overhead. We then suggest methods of handling network partition and network merging situations, and go on to evaluate our approach through simulations. The simulation result shows that our scheme guarantees the uniqueness of allocated address and considerably improves allocation latency and communication overheads.

• Journal of KIISE:Software and Applications
• /
• v.34 no.9
• /
• pp.804-816
• /
• 2007

Sample training data for machine learning often contain irrelevant information or redundant concept. It is also the case that the original data may include noise. If the information collected for constructing learning model is not reliable, it is difficult to obtain accurate information. So the system attempts to find relations or regulations between features and categories in the teaming phase. The feature selection is to remove irrelevant or redundant information before constructing teaming model. for improving its performance. Existing feature selection methods assume that the distribution of documents is balanced in terms of the number of documents for each class and the length of each document. In practice, however, it is difficult not only to prepare a set of documents with almost equal length, but also to define a number of classes with fixed number of document elements. In this paper, we propose a new feature selection method that considers the impurities among the words and unbalanced distribution of documents in categories. We could obtain feature candidates using the word impurity and eventually select the features through unbalanced distribution of documents. We demonstrate that our method performs better than other existing methods via some experiments.

• Journal of KIISE:Computing Practices and Letters
• /
• v.7 no.2
• /
• pp.121-131
• /
• 2001

The failure of network and/or node on which server object is executed is a single point of system failure in the CORBA application. One of the possible ways to overcome such problem is to replicate server objects to several independent nodes. The replicated objects executing same tasks are called object group. In order to provide fault tolerance of server object, this paper proposes and implements new CORBA model that supports the object group based on active replication. The proposed model not only provides interoperability with existing CORBA application but also minimizes additional application interface required to support object group because it uses nop to exchange messages between client and server. And this paper extends IDL structure. Depending to application logic, it makes possible to prevent performance degradation caused by consistency maintenance. At present, this paper supports only active replication. But it can be easily extended to provide warm ancVor cold passive replication without modification of architecture required for active replication.

• Journal of Korean Society for Geospatial Information Science
• /
• v.11 no.3 s.26
• /
• pp.13-21
• /
• 2003

Although more than 90% of the yearly freight has been made through shipping the positioning system for the ships and cargos to support effective transportation of the freights has not been established yet. The government has been working on developing such a positioning system in many perspectives. The main problem in building a such system is that individual shipping companies are running different types of system and databases. This would cause duplicated investment with wasting money and time and this becomes an obstacle to the easier information sharing. Therefore, this study has mainly demonstrated developing a manager's system in the positioning system for the ships and cargos based on client-server scheme using electronic nautical charts on which the international standards are already available. In addition, this study also proposed an independent system which can be used in the ports with different locations thereby preventing duplicated investments.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.48 no.5
• /
• pp.33-41
• /
• 2011

In this paper, we propose a lossless compression scheme for cloud point data of the target construction by using doubleness and decreasing useless information of cloud point data. We use Hough transform to find the horizontal angle between construction and terrestrial LIDAR. This angle is used for the rotation of the cloud point data. The cloud point data can be parallel to x-axis, then y-axis doubleness is increased. Therefore, the cloud point data can be more compressed. In addition, we apply two methods to decrease the number of cloud point data for useless information of them. One is decimation of the cloud point data, the other is to extract the range of y-coordinates of target construction, and then extract the cloud point data existing in the range only. The experimental result shows the performance of proposed scheme. To compress the data, we use only the position information without additional information. Therefore, this scheme can increase processing speed of the compression algorithm.

• Journal of Broadcast Engineering
• /
• v.13 no.2
• /
• pp.245-250
• /
• 2008

Efficient service discovery mechanism is a crucial feature for the usability of a wireless ad-hoc network. A wireless ad-hoc network is a temporal network formed by a collection of wireless mobile nodes without the aid of any existing network infrastructure or centralized administration. We propose an efficient service discovery mechanism using non-disjoint multi-path routing protocol for a wireless ad-hoc network. Our scheme has advantages of not only multi-path routing protocol but also cross-layer service discovery. By simulation, we showed that faster route recovery is possible by maintaining multiple routing paths in each node, and the route maintenance overhead can be reduced by limiting the number of multiple routing paths and by maintaining link/node non-disjoint multi-path.

• Journal of KIISE:Computing Practices and Letters
• /
• v.15 no.9
• /
• pp.695-699
• /
• 2009

In this paper, we propose an automatic prostate segmentation technique using image intensity and gradient information. Our method is composed of four steps. First, rays at regular intervals are generated. To minimize the effect of noise, the start and end positions of the ray are calculated. Second, the profiles on each ray are sorted based on the gradient. And priorities are applied to the sorted gradient in the profile. Third, boundary points are extracted by using gradient priority and intensity distribution. Finally, to reduce the error, the extracted boundary points are corrected by using B-spline interpolation. For accuracy evaluation, the average distance differences and overlapping region ratio between results of manual and automatic segmentations are calculated. As the experimental results, the average distance difference error and standard deviation were 1.09mm $\pm0.20mm$. And the overlapping region ratio was 92%.

• Journal of KIISE:Computer Systems and Theory
• /
• v.30 no.10
• /
• pp.595-602
• /
• 2003

This Paper presents an efficient hybrid techniques for a standard cell placement. The prototype tool adopts a middle-down methodology in which an n${\times}$m grid is imposed over the layout area and cells are assigned to bins forming a global placement. The optimization technique applied in this phase is based on the Relaxation-Based Local Search (RBLS) framework [12］in which a combinatorial search mechanism is driven by an analytical engine. This enables a more global view of the problem and results in complex modifications of the placement in a single search“move.”Details of this approach including a novel placement legalization procedure are presented. When a global placement converges, a detailed placement is formed and further optimized by the optimal interleaving technique[13］. Experimental results on MCNC benchmarking circuits are presented and compared with the Feng Shui's results in［14］. Solution Qualifies are almost the same as the Feng Shui's results.

• Journal of the Institute of Electronics Engineers of Korea TE
• /
• v.37 no.2
• /
• pp.33-40
• /
• 2000

In this study, a shape decomposition method using morphological operations is studied for decomposing the complex shape in 2-D image into its simple primitive elements. The serious drawback of conventional shape representation algorithm is that primitive elements are extracted too much to represent and to describe the shape. To solve these problems, a new shape decomposition algorithm using primitive elements that are similar to the geometrical characteristics of shape and 4 scan modes is proposed in this study. The multiple primitive elements as circle, square, and rhombus are extracted by using multiscan modes in a new algorithm. This algorithm have the characteristics that description error and number of primitive elements is reduced. Then, description efficiency is improved. The procedures is also simple and the processing time is reduced.