• Proceedings of the IEEK Conference
• /
• 2000.07b
• /
• pp.1041-1044
• /
• 2000

The enhancement to the back-propagation algorithm presented in this paper has resulted from the need to extract sparsely connected networks from networks employing product terms. The enhancement works in conjunction with the back-propagation weight update process, so that the actions of weight zeroing and weight stimulation enhance each other. It is shown that the error measure, can also be interpreted as rate of weight change (as opposed to ${\Delta}W_{ij}$), and consequently used to determine when weights have reached a stable state. Weights judged to be stable are then compared to a zero weight threshold. Should they fall below this threshold, then the weight in question is zeroed. Simulation of such a system is shown to return improved learning rates and reduce network connection requirements, with respect to the optimal network solution, trained using the normal back-propagation algorithm for Multi-Layer Perceptron (MLP), Higher Order Neural Network (HONN) and Sigma-Pi networks.

• The Journal of the Acoustical Society of Korea
• /
• v.13 no.1E
• /
• pp.25-36
• /
• 1994

A problem of making a neural network learning self-adaptive to the training set supplied is addressed in this paper. This arises from the aspect in choice of an adequate stepsize for the update of the current weigh vectors according to the training pairs. Related issues in this attempt are raised and fundamentals in neural network learning are introduced. In comparison to the most popular back-propagation scheme, the usefulness and superiority of the proposed weight update algorithm are illustrated by examing the identification of unknown nonlinear systems only from measurements.

• Proceedings of the IEEK Conference
• /
• 2002.07c
• /
• pp.1859-1862
• /
• 2002

The replication of data is used to increase its availability, improve the performance of a system, and advance the fault-tolerance of a system. In this paper, it is required for the information about the location of a primary site of the replicas of each data item. The replicas of each data item are hierarchically organized to a tree based on the fact that the root is the primary replica in partially replicated databases. It eliminates useless propagation since the propagation can be done to only sites having replicas following the hierarchy of data. And our algorithm schedules transactions so that the execution order of updates at each primary site is identical at all sites by using timestamp. Using our algorithm, the consistent data are supplied and the performance of read-only transactions can be improved by using tree structure of replicas of each data item.

• Journal of KIISE:Information Networking
• /
• v.35 no.1
• /
• pp.1-10
• /
• 2008

Once the sensor node in wireless sensor networks is installed, it usually operates without human intervention for a long time. The remote code update scheme is required because it is difficult to recall the sensor node in many situations. Therefore, studies on the reliable and efficient transport protocol for code propagation in wireless sensor networks have been increasingly done. However, by considering only the stability aspect of transmission, most of previous works ignore the consideration on the fast code propagation. This results the energy inefficiency by consuming unnecessary energy due to the slow code propagation. In this paper, in order to overcome limitation of the previous code propagation protocols, we propose a new code propagation protocol called "FCPP(Fast Code Propagation Protocol)". The FCPP aims at improving the reliability at well as performance. For this purpose, the FCPP accomplishes the fast code propagation by using the RTT-based transmission rate control and NACK suppression scheme, which provides a better the network utilization and avoids a unnecessary transmission delay. Based on the ns-2 simulation result, we prove that the FCPP Improves significantly both reliability and performance.

• The KIPS Transactions:PartD
• /
• v.13D no.5 s.108
• /
• pp.641-650
• /
• 2006

XML has the characteristics of self-describing and uses DTD or XML schema in order to constraint its structure. Even though the XML schema is only at the stage of recommendation yet, it will be prevalently used because DTD is not itself XML and has the limitation on the expression power. The structure defined by the XML schema as well as the data of the XML documents can vary due to complex reasons. Those reasons are errors in the XML schema design, new requirements due to new applications, etc. Thus, we propose XML schema evolution operators that are extracted from the analysis of the XML schema updates. These schema evolution operators enable the XML schema updates that would have been impossible without supporting tools if there are a large number of XML documents complying the U schema. In addition, these operators includes the function of automatically finding the update place in the XML documents which are registered to the XSE system, and maintaining the XML documents valid to the XML schema rather than merely well-formed. This paper is the first attempt to update XML schemas of the XML documents and provides the comprehensive set of schema updating operations. Our work is necessary for the XML application development and maintenance in that it helps to update the structure of the XML documents as well as the data in the easy and precise manner.

• Journal of KIISE:Databases
• /
• v.31 no.3
• /
• pp.274-286
• /
• 2004

In mobile client-server environments, mobile clients usually are disconnected with their server because of high cost of wireless communication and keep their own local copies to provide efficient updating the cached map. The update of the server database leads to invalidation of the cached map in the client side. To solve the issues of invalidation of the cached map, it is not efficient to resend part of the updated server database to clients whenever the updating of the server database occurs. This paper proposes a log-based update propagation method to propagate the server's update into its relevant clients by using only update logs. Too many logs increasingly accumulate as the sever database is updated several times. The sequential search of the relevant log data for a specific client is time-consuming. Sending of unnecessary logs should be avoided for reducing the overhead of communication.'re solve these problems, we first define unnecessary logs and then suggest log reduction methods to avoid or cancel creating unnecessary logs. The update log index is used for quickly retrieving relevant logs.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.39 no.6
• /
• pp.591-601
• /
• 2002

Motion compensation is a powerful method to compress an image sequence. Its main drawback is that once an error is occurred, the error propagates through the frames. Recently, the intra update method was proposed to stop the error propagation at the expense of reduction in compression efficiency. This paper proposes an intra update method based on a rate-distortion optimization in error prone environments. The rate and the distortion are estimated using the Lagrangian optimization to select the coding mode and the quantization step size. The proposed method is applied to MPEG-4 codec, and the experimental results show that it is robust to the error such as packet losses comparing with the conventional ones.

• The Journal of Information Technology
• /
• v.5 no.2
• /
• pp.55-67
• /
• 2002

This paper presents an adaptive back propagation algorithm that update the learning parameter by the generated error, adaptively and varies the number of hidden layer node. This algorithm is expected to escaping from the local minimum and make the best environment for convergence to be change the number of hidden layer node. On the simulation tested this algorithm on two learning pattern. One was exclusive-OR learning and the other was $7{\times}5$ dot alphabetic font learning. In both examples, the probability of becoming trapped in local minimum was reduce. Furthermore, in alphabetic font learning, the neural network enhanced to learning efficient about 41.56%~58.28% for the conventional back propagation. and HNAD(Hidden Node Adding and Deleting) algorithm.

• Proceedings of the KIEE Conference
• /
• 1999.07b
• /
• pp.822-825
• /
• 1999

This paper compares two methods of GPS/INS integration ; tightly-coupled integration ana loosely-coupled integration. In the tightly -coupled method an integrated Kalman filter is designed to process raw GPS measurement data for state update and INS data for propagation. The loosely-coupled integration method uses the solution outputs from a stand-alone GPS receiver for update. The loosely-coupled method is simpler and can readily be applied to off-the-self receivers and sensors while the tightly-coupled integration requires access to raw measurement mechanism of the receiver. Simulation result show that the tightly-coupled integration system exhibits better performance and robustness than loosely-coupled integration method.

• Journal of The Korean Astronomical Society
• /
• v.34 no.4
• /
• pp.237-241
• /
• 2001

Our goal is to present a simple volume-of-fluid type interface-tracking algorithm to compressible two-phase flow in two space dimensions. The algorithm uses a uniform underlying Cartesian grid with some cells cut by the tracked interfaces into two subcells. A volume-moving procedure that consists of two basic steps: (1) the update of volume fractions in each grid cell at the end of the time step, and (2) the reconstruction of interfaces from discrete set of volume fractions, is employed to follow the dynamical behavior of the interface motion. As in the previous work with a surface-tracking procedure for general front tracking (LeVeque & Shyue 1995, 1996), a high resolution finite volume method is then applied on the resulting slightly nonuniform grid to update all the cell values, while the stability of the method is maintained by using a large time step wave propagation approach even in the presence of small cells and the use of a time step with respect to the uniform grid cells. A sample preliminary numerical result for an underwater explosion problem is shown to demonstrate the feasibility of the algorithm for practical problems.