• Annual Conference of KIPS
• /
• 2024.05a
• /
• pp.15-17
• /
• 2024

최근 다양한 연구 분야에서는 CUDA 프레임워크를 이용하여 병렬 처리를 통해 연산 시간을 단축하는데 성공하고 있다. 이 중 숄레스키 분해는 양의 정부호 행렬을 하삼각행렬로 분해하는 과정에서 많은 행렬 곱셈이 요구되어 GPU 의 구조적 특징을 활용하면 상당한 가속화가 가능하다. 따라서 이 논문에서는 CUDA 코어에 연산을 할당할 때, 핵심 요소인 블록의 개수와 블록 당 쓰레드 개수를 조절할 수 있는 병렬 숄레스키 분해 연산 프로그램을 구현하였다. 서로 다른 세 종류의 행렬 크기에 대해 다양한 블록 수-쓰레드 수 환경을 설정하여 가속화 정도를 측정한 결과, 각 행렬 별 최적 환경에서 동일 그룹 내 최장 시간 대비, 1000x1000 행렬에서는 약 1.80 배, 2000x2000 행렬에서는 약 2.94 배의 추가적인 가속화를 달성하였다.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.37C no.11
• /
• pp.1077-1083
• /
• 2012

The jamming countermeasures in GNSS includes anti-jamming technique and jammer localization technique. In both techniques, direction of jamming signal is important and generally the MUSIC algorithm is used to find the direction of jamming signal. The MUSIC is super-resolution algorithm for detecting incident direction of signal. But, the search time of MUSIC algorithm is too long because all candidates of incidence angle are searched. This paper proposes the new method that has less computational burdens and therefore faster than the conventional MUSIC algorithm. The proposed method improves performance speed by reducing unnecessary calculations. In the proposed method, the cost function of conventional MUSIC algorithm is decomposed into the sum of squares and if the partial sum of cost function is larger than the minimum cost function so far, then the candidate is rejected and next candidates are searched. If the computed cost function is less than the minimum cost function so far, the minimum cost function so far is replaced with newly computed value. The performance of the proposed method was compared with the conventional MUSIC algorithm using the simulation. The accuracy of the estimaed direction of jamming signal was same as the conventional MUSIC while the search speed of the proposed method was 1.15 times faster than the conventional MUSIC.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.29 no.1A
• /
• pp.31-43
• /
• 2009

In this study, the numerical method is presented, which can consider the various train types and can solve the equations of motion for a vehicle-bridge interaction analysis by non-iteration procedure through formulating the coupled equations of motion. The coupled equations of motion for the vehicle-bridge interaction are solved by the Newmark ${\beta}$ of a direct integration method, and by composing the effective stiffness matrix and the effective force vector according to a analysis step, those can be solved with the same manner of the solving procedure of equilibrium equations in static analysis. Also, the effective stiffness matrix is reconstructed by the Skyline method for increasing the analysis effectiveness. The Cholesky's matrix decomposition scheme is applied to the analysis procedure for minimizing the numerical errors that can be generated in directly calculating the inverse matrix. The equations of motion for the conventional trains are derived, and the numerical models of the conventional trains are idealized by a set of linear springs and dashpots with 16 degrees of freedom. The bridge models are simplified by the 3 dimensional space frame element which is based on the Euler-Bernoulli theory. The rail irregularities of vertical and lateral directions are generated by the PSD functions of the Federal Railroad Administration (FRA). The results of the vehicle-bridge interaction analysis are verified by the experimental results for the railway plate girder bridges of a span length with 12 m, 18 m, and the experimental and analytical data are applied to the low pass filtering scheme, and the basis frequency of the filtering is a 2 times of the 1st fundamental frequency of a bridge bending.

• The Korean Journal of Applied Statistics
• /
• v.26 no.5
• /
• pp.747-758
• /
• 2013

The covariance matrix is important in multivariate statistical analysis and a sample covariance matrix is used as an estimator of the covariance matrix. High dimensional data has a larger dimension than the sample size; therefore, the sample covariance matrix may not be suitable since it is known to perform poorly and event not invertible. A number of covariance matrix estimators have been recently proposed with three different approaches of shrinkage, thresholding, and modified Cholesky decomposition. We compare the performance of these newly proposed estimators in various situations.

• Journal of Electrical Engineering and Technology
• /
• v.9 no.6
• /
• pp.1832-1842
• /
• 2014

A novel probabilistic small signal stability analysis (PSSSA) method considering load correlation is proposed in this paper. The superiority Latin hypercube sampling (LHS) technique combined with Monte Carlo simulation (MCS) is utilized to investigate the probabilistic small signal stability of power system in presence of load correlation. LHS helps to reduce the sampling size, meanwhile guarantees the accuracy and robustness of the solutions. The correlation coefficient matrix is adopted to represent the correlations between loads. Simulation results of the two-area, four-machine system prove that the proposed method is an efficient and robust sampling method. Simulation results of the 16-machine, 68-bus test system indicate that load correlation has a significant impact on the probabilistic analysis result of the critical oscillation mode under a certain degree of load uncertainty.

• Wind and Structures
• /
• v.9 no.3
• /
• pp.231-243
• /
• 2006

An improvement to the spectral representation algorithm for the simulation of wind velocity fields on large scale structures is proposed in this paper. The method proposed by Deodatis (1996) serves as the basis of the improved algorithm. Firstly, an interpolation approximation is introduced to simplify the computation of the lower triangular matrix with the Cholesky decomposition of the cross-spectral density (CSD) matrix, since each element of the triangular matrix varies continuously with the wind spectra frequency. Fast Fourier Transform (FFT) technique is used to further enhance the efficiency of computation. Secondly, as an alternative spectral representation, the vectors of the triangular matrix in the Deodatis formula are replaced using an appropriate number of eigenvectors with the spectral decomposition of the CSD matrix. Lastly, a turbulent wind velocity field through a vertical plane on a long-span bridge (span-wise) is simulated to illustrate the proposed schemes. It is noted that the proposed schemes require less computer memory and are more efficiently simulated than that obtained using the existing traditional method. Furthermore, the reliability of the interpolation approximation in the simulation of wind velocity field is confirmed.

• Journal of Mechanical Science and Technology
• /
• v.15 no.8
• /
• pp.1090-1096
• /
• 2001

In this paper, new methods for efficiently solving linear acceleration equations of multibody dynamic simulation exploiting sparsity for real-time simulation are presented. The coefficient matrix of the equations tends to have a large number of zero entries according to the relative joint coordinate numbering. By adequate joint coordinate numbering, the matrix has minimum off-diagonal terms and a block pattern of non-zero entries and can be solved efficiently. The proposed methods, using sparse Cholesky method and recursive block mass matrix method, take advantages of both the special structure and the sparsity of the coefficient matrix to reduce computation time. The first method solves the η$\times$η sparse coefficient matrix for the accelerations, where η denotes the number of relative coordinates. In the second method, for vehicle dynamic simulation, simple manipulations bring the original problem of dimension η$\times$η to an equivalent problem of dimension 6$\times$6 to be solved for the accelerations of a vehicle chassis. For vehicle dynamic simulation, the proposed solution methods are proved to be more efficient than the classical approaches using reduced Lagrangian multiplier method. With the methods computation time for real-time vehicle dynamic simulation can be reduced up to 14 per cent compared to the classical approach.

• Geomechanics and Engineering
• /
• v.3 no.3
• /
• pp.169-188
• /
• 2011

Response of buried flexible pipe-soil system is studied, through numerical analysis, with respect to deflection and buckling in a spatially varying soil media. In numerical modeling procedure, soil parameters are modeled as two-dimensional non-Gaussian homogeneous random field using Cholesky decomposition technique. Numerical analysis is performed using random field theory combined with finite difference numerical code FLAC 5.0 (2D). Monte Carlo simulations are performed to obtain the statistics, i.e., mean and variance of deflection and circumferential (buckling) stresses of buried flexible pipe-soil system in a spatially varying soil media. Results are compared and discussed in the light of available analytical solutions as well as conventional numerical procedures in which soil parameters are considered as uniformly constant. The statistical information obtained from Monte Carlo simulations is further utilized for the reliability analysis of buried flexible pipe-soil system with respect to deflection and buckling. The results of the reliability analysis clearly demonstrate the influence of extent of variation and spatial correlation structure of soil parameters on the performance assessment of buried flexible pipe-soil systems, which is not well captured in conventional procedures.

• Geomechanics and Engineering
• /
• v.23 no.1
• /
• pp.15-30
• /
• 2020

The probabilistic bearing capacity of a strip footing placed on the edge of a purely cohesive reinforced soil slope is computed by combining lower bound finite element limit analysis technique with random field method and Monte Carlo simulation technique. To simulate actual field condition, anisotropic random field model of undrained soil shear strength is generated by using the Cholesky-Decomposition method. With the inclusion of a single layer of reinforcement, dimensionless bearing capacity factor, N always increases in both deterministic and probabilistic analysis. As the coefficient of variation of the undrained soil shear strength increases, the mean N value in both unreinforced and reinforced slopes reduces for particular values of correlation length in horizontal and vertical directions. For smaller correlation lengths, the mean N value of unreinforced and reinforced slopes is always lower than the deterministic solutions. However, with the increment in the correlation lengths, this difference reduces and at a higher correlation length, both the deterministic and probabilistic mean values become almost equal. Providing reinforcement under footing subjected to eccentric load is found to be an efficient solution. However, both the deterministic and probabilistic bearing capacity for unreinforced and reinforced slopes reduces with the consideration of loading eccentricity.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.7
• /
• pp.3599-3619
• /
• 2019

In human activity recognition system both static and motion information play crucial role for efficient and competitive results. Most of the existing methods are insufficient to extract video features and unable to investigate the level of contribution of both (Static and Motion) components. Our work highlights this problem and proposes Static-Motion fused features descriptor (SMFD), which intelligently leverages both static and motion features in the form of descriptor. First, static features are learned by two-stream 3D convolutional neural network. Second, trajectories are extracted by tracking key points and only those trajectories have been selected which are located in central region of the original video frame in order to to reduce irrelevant background trajectories as well computational complexity. Then, shape and motion descriptors are obtained along with key points by using SIFT flow. Next, cholesky transformation is introduced to fuse static and motion feature vectors to guarantee the equal contribution of all descriptors. Finally, Long Short-Term Memory (LSTM) network is utilized to discover long-term temporal dependencies and final prediction. To confirm the effectiveness of the proposed approach, extensive experiments have been conducted on three well-known datasets i.e. UCF101, HMDB51 and YouTube. Findings shows that the resulting recognition system is on par with state-of-the-art methods.