• Journal of KIISE
• /
• v.44 no.12
• /
• pp.1290-1295
• /
• 2017

Document classification, a task of classifying the category of each document based on text, is one of the fundamental areas for natural language processing. Document classification may be used in various fields such as topic classification and sentiment classification. Neural network models for document classification can be divided into two categories: word-level models and character-level models that treat words and characters as basic units respectively. In this study, we propose a neural network model that combines character-level and word-level models to improve performance of document classification. The proposed model extracts the feature vector of each word by combining information obtained from a word embedding matrix and information encoded by a character-level neural network. Based on feature vectors of words, the model classifies documents with a hierarchical structure wherein recurrent neural networks with attention mechanisms are used for both the word and the sentence levels. Experiments on real life datasets demonstrate effectiveness of our proposed model.

• Proceedings of the Korean Society for Agricultural Machinery Conference
• /
• 2017.04a
• /
• pp.167-167
• /
• 2017

스마트팜 내부의 3차원 공간의 온도, 습도, 기압, 공기질 분석을 통한 돈사 미세 조절 기술에 대한 연구가 진행 중이다. 해당 특성 중에서 기압을 제외한 환경인자들은 돈사 내의 구조 특성상 위치별로, 시간별로 매우 상이한 변이의 형태를 보인다. 일정 시점을 기준으로 계측 지점 이외의 지점에 대한 환경인자들을 공간적으로 추정하는 기술은 대표적으로 컴퓨터 분석 기술에 의존하고 있다. 시간 복잡도가 매우 높은 CFD(Computer Fluid Dynamics) 방식은 정밀도 측면에서 유리하나, 상응하는 제어 기술/하드웨어 등의 부재로 모델링 결과의 활용도가 낮다고 볼 수 있다. 본 연구에서는 CFD를 수행하는 과정에 있어 실질적으로 유효한 단위로 공간 분해능을 낮추고, 동등한 크기의 세부 공간에 대한 모델링을 병렬적으로 수행하기 위한 방안을 연구하였다. 실험적으로 돈사 환경을 3차원으로 구성하기 위하여, 공기 흡입구, 배출구, 기둥, 덕트 요소를 포함시켰다. 실내 공간을 1차적으로 가로, 세로, 높이방향으로 $3{\times}3{\times}3$ 균등 분배한 후 3차원 행렬로 분할하였다. 각 분할된 행렬에 대한 연산 수행을 위하여 현재까지 대중에 공개된 SBC(Single Board Computer) 중 가장 높은 연산 수행 능력이 있는 Odroid-XU4(Hardkernel, AnYang, Korea) 16식을 병렬 클러스터링 기술로 연동하였다. 하나의 AP당 8개의 코어가 내장되어 있으므로, 총 128개의 코어를 이용하여 동시에 128개의 3D 정방행렬 연산이 가능하도록 구성하였다. 모델링을 위한 수학적 모델로는 실험적으로 Steady turbulent model (Newtonian coefficient)을 이용하였다. 클러스터링을 이용한 병렬 처리의 특성상 균등한 정보량을 동시에 배분해야 하므로 108 ($27{\times}4$)개의 코어를 이용하여 1차적으로 나뉜 공간을 다시 4등분하여 동시에 $12{\times}12{\times}12$에 해당하는 공간 분해능에 대한 처리를 동시에 수행할 수 있도록 하였다. 2단계에 걸쳐 분할한 공간 세그먼트에 대한 클러스터링 연산 수행 결과 초당 15회 정도의 연산을 수행할 수 있었으며, 시간 분해능을 100으로 설정한 경우 약 5초가 수행되었다. 선행적으로 수행하였던 CFD 모델링 (OpenFOAM)과 비교하였을 때 상대적으로 정밀도가 낮은 3차원 모델링 결과를 얻을 수 있었다. 모델링에 소요되는 시간을 비약적으로 경감 시킨 장점을 살려 적정한 공간 분할 기법과 추가로 발생하는 다수의 바운더리 조건을 근사적으로 추정할 수 있는 데이터 마이닝 기술이 보완되어야 할 것이다.

• The Korean Journal of Applied Statistics
• /
• v.29 no.4
• /
• pp.581-594
• /
• 2016

We introduce innovations state space exponential smoothing models (ISS-ESM) that can analyze time series with multiple seasonal patterns. Especially, in order to control complex structure existing in the multiple patterns, the model equations use a matrix consisting of seasonal updating parameters. It enables us to group the seasonal parameters according to their similarity. Because of the grouped parameters, we can accomplish the principle of parsimony. Further, the ISS-ESM can potentially accommodate any number of multiple seasonal patterns. The models are applied to predict electricity demand in Korea that is observed on hourly basis, and we compare their performance with that of the traditional exponential smoothing methods. It is observed that the ISS-ESM are superior to the traditional methods in terms of the prediction and the interpretability of seasonal patterns.

• Journal of the Korea Society of Computer and Information
• /
• v.5 no.3
• /
• pp.29-32
• /
• 2000

Layered video coding schemes allow the video information to be transmitted in multiple video bitstreams to achieve scalability. they are attractive in theory for two reasons. First, they naturally allow for heterogeneity in networks and receivers in terms of client processing capability and network bandwidth. Second, they correspond to optimal utilization of available bandwidth when several video qualify levels are desired. In this paper we propose a scalable video codec architectures with motion estimation, which is suitable for real-time audio and video communication over packet networks. The coding algorithm is compatible with ITU-T recommendation H.263+ and includes various techniques to reduce complexity. Fast motion estimation is Performed at the H.263-compatible base layer and used at higher layers, and perceptual macroblock skipping is performed at all layers before motion estimation. Error propagation from packet loss is avoided by Periodically rebuilding a valid Predictor in Intra mode at each layer.

• Journal of the Institute of Electronics Engineers of Korea TC
• /
• v.49 no.5
• /
• pp.37-43
• /
• 2012

The closed-loop multiple-input multiple-output (MIMO) system has been adopted by long term evolution (LTE) system. Many techniques are proposed to enhance the transmission of LTE's advanced version to meet the increasing requirement, in which differential codebook gains a lot of interest. Previous researches on designing differential codebooks focused on quasi-diagonal unitary matrix which cannot guarantee the equal gain property. The equal gain property is very important to uplink because the performance of uplink is very sensitive to the peak-to-average power ratio (PAPR). In this paper, we derive the analytical expression of average bit error rate and PAPR for differential precoding MIMO system. Using the analytical results, we investigate the performances of several differential precoding schemes considering non-linear amplifier at the transmitter. Some selected simulation results indicate that the conventional differential precoding schemes have good performances without the consideration of non-linear amplifier. While considering non-linear amplifier, the proposed differential codebook outperforms other differential precoding schemes because it maintains the equal gain per transmit antenna.

• Journal of Korean Society of Steel Construction
• /
• v.23 no.4
• /
• pp.429-438
• /
• 2011

In this study, the explicit numerical algorithm was proposed to simulate the stress erection process and ultimate-load analysis of the strarch (stressed arch) system. The strarch system is a unique and innovative structural system and member prestress comprising prefabricated plane truss frames erected through a post-tensioning stress erection procedure. The flexible bottom chord, which has sleeve and gap details, is closed by the reaction force of the prestressing tendon. The prestress imposed on the tendon will enable the strarch system to be erected. This post-tensioning process is called "stress erection process." During this process, plastic rigid-body rotation occurs to the flexible top chord due to the excessive amount of plastic strain, and the structural characteristic is unstable. In this study, the dynamic relaxation method (DRM) was adopted to calculate the nonlinear equilibrium equation of the system, and a displacement-based finite-element-formulated filament beam element was used to simulate the nonlinear behavior of the top chord sections of the strarch system. The section of the filament beam element was composed by the amount of filaments, which can be modeled by various material models. The Ramberg-Osgood and bilinear kinematic elastic plastic material models were formulated for the nonlinear material behaviors of the filaments. The numerical results that were obtained in the present study were compared with the experiment results of the stress erection and with the results of the ultimate-load analysis of the strarch unit frame. The results of the present studies are in good agreement with the previous experiment results, and the explicit DRM enabled the analysis of the post-buckling behaviors of the strarch unit frame.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2014.10a
• /
• pp.295-298
• /
• 2014

In this paper, an efficient design of HEVC Adaptive Loop Filter(ALF) for filter coefficients estimation is proposed. The ALF performs Cholesky decomposition of $10{\times}10$ matrix iteratively to estimate filter coefficients. The Cholesky decomposition of the ALF consists of root and division operation which is difficult to implement in a hardware design because it needs to many computation rate and processing time due to floating-point unit operation of large values of the Maximum 30bit in a LCU($64{\times}64$). The proposed hardware architecture is implemented by designing a root operation based on Cholesky decomposition by using multiplexer, subtracter and comparator. In addition, The proposed hardware architecture of efficient and low computation rate is implemented by designing a pipeline architecture using characteristic operation steps of Cholesky decomposition. An implemented hardware is designed using Xilinx ISE 14.3 Vertex-6 XC6VCX240T FPGA device and can support a frame rate of 40 4K Ultra HD($4096{\times}2160$) frames per second at maximum operation frequency 150MHz.

• Journal of the Korea institute for structural maintenance and inspection
• /
• v.22 no.1
• /
• pp.64-71
• /
• 2018

Project Management Institute of America separates the types of risk with external risks and internal risks. The external risk is an uncontrollable risk in projects such as changes of policy and related systems, climate, natural disasters, exchange rates and so on. The internal risk is an existing risk in the project itself that is controllable items in the project. Technical risks in project management are cost, quality, time, safety and environment. Therefore, both the external and internal risks should be managed to perform the construction project successfully. In particular, we can secure the quality and safety of facilities through the technical risk management. The importance of potential risk management has been emerging as a major interest and the lack of risk management delays projects and increases construction costs with negative effects of the building safety since the complex building, which is composed of a great number of facilities, consists of many project units and there are conflicts between various participants and stake-holders. This study presents the ways of establishing risk management processes to ensure the safety of the complex building. To that end, establishing procedure of risk management processes is presented and types of risk and factors in construction projects and counter strategies are presented as available risk information on the stages.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.22 no.3
• /
• pp.94-100
• /
• 2021

To improve the productivity of cargo containment construction for a membrane LNG carrier, it is important to shorten the installation period and process of the scaffolding system, which is a construction workbench of a cargo containment for a membrane LNG carrier. As an effective method, opinions are being gathered to enlarge the lifting unit from the existing two stages to eight stages. On the other hand, the stresses around the pin and hole will increase significantly because of the increase in lifting load according to the large size of the module. The purpose of this study was to establish a new large module-lifting plan by introducing TRIZ to solve these problems. This study adopted a method to utilize 40 inventive principles, which is one of the various problem-solving tools of TRIZ. First, technical contradictions were derived, the engineering parameters were selected. Second, efficient inventive principles were selected to overcome the technical contradictions using a contradiction matrix. Finally, the general and specific solutions were derived through the selected inventive principle, and structural analysis confirmed that the stress generated in the structure was low. The utility of TRIZ was confirmed by the successful lifting of large modules using the established lifting method.