• Journal of the Korean Data and Information Science Society
• /
• 제27권2호
• /
• pp.549-557
• /
• 2016

In this paper we propose multiclassification method for large data sets by ensembling least squares support vector machines (LS-SVM) with principal components instead of raw input vector. We use the revised one-vs-all method for multiclassification, which is one of voting scheme based on combining several binary classifications. The revised one-vs-all method is performed by using the hat matrix of LS-SVM ensemble, which is obtained by ensembling LS-SVMs trained using each random sample from the whole large training data. The leave-one-out cross validation (CV) function is used for the optimal values of hyper-parameters which affect the performance of multiclass LS-SVM ensemble. We present the generalized cross validation function to reduce computational burden of leave-one-out CV functions. Experimental results from real data sets are then obtained to illustrate the performance of the proposed multiclass LS-SVM ensemble.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2004년도 학술대회지
• /
• pp.503-508
• /
• 2004

This paper proposes an efficient computation method for fixed and mixed polarity Reed -Muller function vector over Galois field GF(p). Function vectors of fixed polarity Heed Muller function with single variable can be generated by proposed method. The n-variable function vectors can be calculated by means of the Kronecker product of a single variable function vector corresponding to each variable. Thus, all fixed and mixed polarity Reed-Muller function vectors are calculated directly without using a polarity function vector table or polarity coefficient matrix.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2008년도 하계종합학술대회
• /
• pp.883-884
• /
• 2008

This paper presents a method for classification of underwater transient signals using, which employs a binary image pattern of the mel-frequency cepstral coefficients(MFCC) as a feature vector and a neural network as a classifier. A feature vector is obtained by taking DCT and 1-bit quantization for the square matrix of the MFCC sequences. The classifier is a feed-forward neural network having one hidden layer and one output layer, and a back propagation algorithm is used to update the weighting vector of each layer. Experimental results with some underwater transient signals demonstrate that the proposed method is very promising for classification of underwater transient signals.

• 제어로봇시스템학회논문지
• /
• 제20권9호
• /
• pp.900-907
• /
• 2014

This paper proposes a MPC (Model Predictive Control) method for the torque and flux controls of induction motor. The proposed MPC method selects the optimized voltage vector for the matrix converter control using the predictive modeling equation of the induction motor and cost function. Hence, the reference voltage vector that minimizes the cost function of the torque and flux error within the control period is selected and applied to the actual system. As a result, it is possible to perform the torque and flux control of induction motor using only the MPC controller without a PI (Proportional-Integral) or hysteresis controller. Even though the proposed control algorithm is more complicated and has lots of computations compared with the conventional MPC, it can perform torque ripple reduction by synthesizing voltage vectors of various magnitude. This feature provides the reduction of amount of calculations and the improvement of the control performance through the adjustment of the number of the unit vectors n. The proposed control method is validated through the PSIM simulation.

• Journal of the Korean Data and Information Science Society
• /
• 제28권4호
• /
• pp.721-731
• /
• 2017

이 연구에서는 Lasso 페널티 방법을 이용한 주성분분석 방법을 소개한다. 주성분분석에 Lasso 페널티를 적용하는 방법으로 흔히 사용되는 방법은 크게 두 가지가 있다. 첫 번째 방법은 주성분을 반응변수로 놓고 원 자료행렬을 설명변수로 하는 회귀분석의 회귀계수를 이용하여 최적의 선형결 합 벡터를 구할 때 Lasso 페널티 (일반적으로 elastic net 페널티)를 부과하는 방법이다. 두 번째 방법은 원자료행렬을 비정칙값 분해로 근사하고 남은 잔차행렬에 Lasso 페널티를 부과하여 최적의 선형결합 벡터를 구하는 방법이다. 이 연구에서는 주성분 분석에 Lasso 페널티를 부과하는 이 두 가지 방법들을 자세하게 개관하는데, 이 방법들은 변수 숫자가 표본크기보다 큰 경우에도 적용가능한 장점이 있다. 또한 실제 자료분석에서 R 프로그램을 통해 두 방법을 적용하고 그 결과를 비교한다. 구체적으로 변수 숫자가 표본크기보다 큰 Ahamad (1967)의 crime 자료에 적용한다.

• Journal of Power Electronics
• /
• 제17권4호
• /
• pp.1048-1057
• /
• 2017

An indirect matrix converter (IMC) is a modern power generation system that enables a direct ac/ac conversion without the need for any bulky and limited lifetime electrolytic capacitor. This system also allows four-quadrant operation, generation of sinusoidal output voltage waveforms with variable frequency and amplitude, and control of input power factor. This study proposes a pulse-width modulation-based sliding-mode controller to achieve unity input-power factor operation of the IMC independently of the active power exchanged with the grid, as well as a fast dynamic response. The designed equivalent control law determines, at each sampling period, the appropriate q-axis component of the modulated input current to be injected into the grid through the LC input filter. An integral term of the error is included in the expression of the sliding surface to increase the accuracy of the control method. A double space vector modulation method is used to synthesize the direction of the space vector of the input currents as required by the sliding-mode controller and the space vectors of the target output voltages. Simulation and experimental results are provided to show the effectiveness and evaluate the performance of the proposed control method.

• 한국전자거래학회:학술대회논문집
• /
• 한국전자거래학회 2005년도 e-Biz World Conference 2005
• /
• pp.130-134
• /
• 2005

This research describes the comparison broker's role and its effectiveness measurement using a developed logical framework of comparison shopping service. And verifies that seller-led comparison challenge method provide comparison information of products to buyers more efficiently. In electronic commerce, buyer's satisfaction of purchase (S) can be defined as an interactive function between seller's competitiveness vector (P) of products that supplied to the market, and buyer's informed level vector (B) of products that is known from a lot of sources. Then the buyer's informed level can be changed through the information analysis among products by transformation process using comparison matrix (C). So the role of comparison shopping is to construct a comparison matrix and to serve it to the buyers, and to change the buyer's informed level. The changed informed level influences a buyer's satisfaction, that improved satisfaction of purchase is defined as the effectiveness of comparison shopping. As a perfect provision and usage of comparison matrix is impossible cause of cognitive limit, the most efficient method for improving the comparison effectiveness is the comparison challenge that detects the comparison elements of the largest buyer's information efficiency, and then to be compared between elementary products selectively. This research verifies the substantial superiority of comparison challenge through television market data experiments.

• Journal of Mechanical Science and Technology
• /
• 제16권8호
• /
• pp.1072-1078
• /
• 2002

The constrained motion requires the determination of constraint force acting on unconstrained systems for satisfying given constraints. Most of the methods to decide the force depend on numerical approaches such that the Lagrange multiplier method, and the other methods need vector analysis or complicated intermediate process. In 1992, Udwadia and Kalaba presented the generalized inverse method to describe the constrained motion as well as to calculate the constraint force. The generalized inverse method has the advantages which do not require any linearization process for the control of nonlinear systems and can explicitly describe the motion of holonomically and/or nongolonomically constrained systems. In this paper, an explicit equation to describe the constrained motion is derived by minimizing the performance index, which is a function of constraint force vector, with respect to the constraint force. At this time, it is shown that the positive-definite weighting matrix in the performance index must be the inverse of mass matrix on the basis of the Gauss's principle and the derived differential equation coincides with the generalized inverse method. The effectiveness of this method is illustrated by means of two numerical applications.

• 한국정보과학회논문지:시스템및이론
• /
• 제32권11_12호
• /
• pp.692-704
• /
• 2005

컴퓨터 하드웨어의 급속한 발전으로 그래픽 프로세서 유닛(Graphics Processor Units : GPUs)은 굉장한 메모리 대역폭과 산술 능역을 보유하게 되어 범용 계산에 많이 활용되고 있으며, 특히 계산 집약적인 물리 기반 시뮬레이션(physics based simulation)의 GPU 구현이 활발하게 연구되고 있다. 물리 기반 시뮬레이션의 기본이 되는 미분방정식 풀이 과정에서 삼중대각행렬(tridiagonal matrix) 시스템은 유한차분(finite-difference) 근사에 의해서 자주 나타나는 선형시스템으로 물리 기반 시뮬레이션 관점에서 삼중대각행렬 시스템의 빠른 풀이는 중요한 연구 분야이다. 본 논문에서는 GPU에서 삼중대각행렬 시스템 풀이를 빠르게 구현할 수 있는 방법을 제안한다. 벡터 프로세서(vector processor) 계산에서 삼중대각행렬 시스템 풀이 방법으로 널리 사용되는 cyclic reduction 또는 odd-even reduction 알고리즘을 GPU에서 구현하였다. 본 논문에서 제안한 방법을 삼중대각행렬 시스템 풀이 방법으로 잘 알려져 있는 Thomas 방법과 GPU를 이용한 선형시스템 풀이에서 좋은 성과를 보이고 있는 conjugate gradient 방법과 비교할 때 상당한 성능 향상을 얻을 수 있었다. 또한, 열전도(heat conduction) 방정식, 이류 확산(advection-diffusion) 방정식, 얕은 물(shallow water) 방정식에 의한 물리 기반 시뮬레이션의 GPU 구현에 본 논문에서 제안한 방법을 사용하여 1024x1024 격자의 계산 영역에서 초당 35프레임 이상의 놀라운 성능을 보여주었다.

• Earthquakes and Structures
• /
• 제3권3_4호
• /
• pp.383-411
• /
• 2012

The objective of the present paper is to review and implement the most recent developments in the Spectral Element Method (SEM), as well as improve aspects of its implementation in the study of wave propagation by numerical simulation in elastic unbounded domains. The classical formulation of the method is reviewed, and the construction of the mass matrix, stiffness matrix and the external force vector is expressed in terms of matrix operations that are familiar to earthquake engineers. To account for the radiation condition at the external boundaries of the domain, a new absorbing boundary condition, based on the Perfectly Matched Layer (PML) is proposed and implemented. The new formulation, referred to as the Multi-Axial Perfectly Matched Layer (M-PML), results from generalizing the classical Perfectly Matched Layer to a medium in which damping profiles are specified in more than one direction.