• KIISE Transactions on Computing Practices
• /
• v.23 no.5
• /
• pp.322-327
• /
• 2017

The classification problem in the field of Natural Language Processing has been studied for a long time. Continuing forward with our previous research, which classifies large-scale text using Convolutional Neural Networks (CNN), we implemented Recurrent Neural Networks (RNN), Long-Short Term Memory (LSTM) and Gated Recurrent Units (GRU). The experiment's result revealed that the performance of classification algorithms was Multinomial Naïve Bayesian Classifier < Support Vector Machine (SVM) < LSTM < CNN < GRU, in order. The result can be interpreted as follows: First, the result of CNN was better than LSTM. Therefore, the text classification problem might be related more to feature extraction problem than to natural language understanding problems. Second, judging from the results the GRU showed better performance in feature extraction than LSTM. Finally, the result that the GRU was better than CNN implies that text classification algorithms should consider feature extraction and sequential information. We presented the results of fine-tuning in deep neural networks to provide some intuition regard natural language processing to future researchers.

• Journal of KIISE:Computing Practices and Letters
• /
• v.16 no.4
• /
• pp.427-431
• /
• 2010

Multidimensional scaling (MDS) is a widely used method for dimensionality reduction, of which purpose is to represent high-dimensional data in a low-dimensional space while preserving distances among objects as much as possible. MDS has mainly been applied to data visualization and feature selection. Among various MDS methods, the classical MDS is not readily applicable to data which has large numbers of objects, on normal desktop computers due to its computational complexity. More precisely, it needs to solve eigenpair problems on dissimilarity matrices based on Euclidean distance. Thus, running time and required memory of the classical MDS highly increase as n (the number of objects) grows up, restricting its use in large-scale domains. In this paper, we propose an efficient approximation algorithm for the classical MDS based on divide-and-conquer and CUDA. Through a set of experiments, we show that our approach is highly efficient and effective for analysis and visualization of data consisting of several thousands of objects.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.10
• /
• pp.190-196
• /
• 2014

This paper proposes a mobile-optimized pedestrian detection method using Cascade of HOG(Histograms of Oriented Gradients) for ADAS(Advanced Driver Assistance System) on smartphones. In order to use the limited resource of mobile platforms efficiently, the method is implemented by the OpenCL(Open Computing Language) library, and its processing time is reduced in the following two aspects. Firstly, the method sets a program build option specifically and adjusts work group sizes as variety of kernels in the host code. Secondly, it utilizes local memory and a LUT(Look-Up Table) in the kernel code to accelerate the program. For performance evaluation, the developed algorithm is compared with the mobile CPU-based OpenCV(Open Computer Vision) for Android function. The experimental results show that the processing speed is 25% faster than the OpenCV hogcascade.

• Journal of KIISE:Information Networking
• /
• v.33 no.2
• /
• pp.165-174
• /
• 2006

Data dissemination using either flooding or legacy ad-hoc routing protocol is not realistic approach in the wireless sensor networks, which are composed of sensor nodes with very weak computing power, small memory and limited battery. In this paper, we propose the ELF(Energy-efficient Localized Flooding) protocol. The ELF is energy-efficient data dissemination protocol for wireless sensor networks. In the ELF protocol, there are two data delivery phases between fixed source and mobile sink node. The first phase, before the tracking zone, sensing data are forwarded by unicasting. After that, within the tracking zone, sensing data are delivered by localized flooding. Namely, the ELF Properly combines advantages from both unicasting and flooding. According to evaluation results by simulation, the proposed ELF protocol maintains very high data delivery ratio with using a little energy. Also, the property of average delay is better than others. From our research results, the ELF is very effective data dissemination protocol for wireless sensor networks.

• Journal of Internet Computing and Services
• /
• v.9 no.6
• /
• pp.117-125
• /
• 2008

With increase of concern about the Ubiquitous application, the necessity of the computer system which is miniaturized is becoming larger. The ARM processor is showing a high share from embedded system market. In this paper, ideal method for RTD-1000 controller construction and development is described using ARM microcontroller. Existing RTD-1000 measures distance of disconnection or defect of sensing casket by measuring receiving reflected wave which was sent via copper wire inside the leaking sensing rod. Using this RTD-1000, leakage and breakage of water and oil pipe can be sensed and it reports damage results to the networks. But, existing RTD-1000 wastes hardware resources much and costs a great deal to installation. Also, it needs a cooling device because the heating problem, and has some problem of the secondary memory unit such as the hard disk. So, long tenn maintenance has some problems in the outside install place. In this paper, for the resolving the problem of RTD-1000, RTD-1000A embedded system based on ARM is proposed and simulated.

• Journal of Internet Computing and Services
• /
• v.12 no.5
• /
• pp.63-72
• /
• 2011

Binary translation is a kind of the emulation method which converts a binary code compiled on the particular instruction set architecture to the new binary code that can be run on another one. It has been mostly used for migrating legacy systems to new architecture. In recent, binary translation is used for instrumenting programs without modifying source code, because it enables inserting additional codes dynamically, For general application, there already exists some instrumentation software using binary translation, such as dynamic binary analyzers and virtual machine monitors. On the other hand, in order to be benefited from binary translation in kernel-level, a few issues, which include system performance, memory management, privileged instructions, and synchronization, should be treated. These matters are derived from the structure of the kernel, and the difference between the kernel and user-level application. In this paper, we present a scheme to apply binary translation and dynamic instrumentation on kernel. We implement it on Linux kernel and demonstrate that kernel-level binary translation adds an insignificant overhead to performance of the system.

• Journal of KIISE:Computing Practices and Letters
• /
• v.12 no.4
• /
• pp.264-274
• /
• 2006

Recently, a variety of organizations including enterprises tend to try to use reporting tools as a data analysis tool for decision making support because reporting tools are capable of formatting data flexibly. Traditional reporting tools have thin-client structure in which all of dynamic documents are generated in the server side. This structure enables reporting tools to avoid repetitive process to generate dynamic documents, when many clients intend to access the same dynamic document. However, generating dynamic documents for data analysis doesn't consider a number of potential readers and increases requests to the server by making clients input various parameters at short intervals. In the structure of the traditional reporting tools, the increase of these requests leads to the increase of processing load in the server side. Thus, we present the reporting tool that can generate dynamic documents at the client side. This reporting tool has a processing mechanism to deal with a number of data despite the limited memory capacity of the client side.

• The KIPS Transactions:PartB
• /
• v.8B no.6
• /
• pp.625-632
• /
• 2001

The electronic stores have realized that they need to understand their customers and to quickly response their wants and needs. To be successful in increasingly competitive Internet marketplace, recommender systems are adapting data mining techniques. One of most successful recommender technologies is collaborative filtering (CF) algorithm which recommends products to a target customer based on the information of other customers and employ statistical techniques to find a set of customers known as neighbors. However, the application of the systems, however, is not very suitable for seasonal products which are sensitive to time or season such as refrigerator or seasonal clothes. In this paper, we propose a new adjusted item-based recommendation generation algorithms called the exponentially weighted collaborative filtering recommendation (EWCFR) one that computes item-item similarities regarding seasonal products. Finally, we suggest the recommendation system with relatively high quality computing time on main memory database (MMDB) in XML since the collaborative filtering systems are needed that can quickly produce high quality recommendations with very large-scale problems.

• Journal of the Korea Society of Computer and Information
• /
• v.15 no.8
• /
• pp.107-115
• /
• 2010

RFID (Radio Frequency IDentification) technology, automatic identification and data capture technologies in ubiquitous computing is an essential skill. Low-cost Radio Frequency Identification tags using memory and no physical contact due to the ease of use and maintenance of excellence are going to use expanded. However, it is possible to the illegal acquisition of the information between RFID tags and readers because RFID uses the RF signal, and the obtained information can be used for the purpose of location tracking and invasion of privacy. In this paper, we proposed the security scheme to protect against the illegal user location tracking and invasion of privacy. The security scheme proposed in this paper, using Gray Code and reduced the capacity of the calculation of the actual tags, However, it is impossible for the malicious attacker to track information because tag information transmitted from the reader is not fixed. Therefore, even if the tags information is obtained by a malicious way, our scheme provides more simple and safe user privacy than any other protection methods to protect user privacy, because not actual information but encrypted information is becoming exposed.

• The Journal of Korean Association of Computer Education
• /
• v.18 no.2
• /
• pp.83-90
• /
• 2015

In an virtualization environment, several virtual machines use physical resources together. If a specific virtual machine uses to much of the computing resources, other machines may not be working properly. There are various method to solve this problem. Most representative study is to migrate a specified virtual machines to a different server, a target server. In this study, server load can be transferred to a target server by the remigrate of the load imposed on virtual machine. It is still problematic that virtual machine has to remigrate to a different server. This thesis has proposed the algorithm determining the remigration targets by applying dynamic thresholds to solve those problems. The migration algorithm applies dynamic thresholds according to the following criteria. Firstly, the usage of CPU, network and memory; secondly, decide the set of artificial machine and the target server based on the resources surpassed thresholds; thirdly, determine artificial machines based on the resource usage in the target server.