• The KIPS Transactions:PartA
• /
• v.8A no.4
• /
• pp.439-446
• /
• 2001

The main stream of parallel programming today is using synchronous algorithms, where processor synchronization for correct computation and workload balance are essential. Overall performance of the whole system is dependent upon the performance of the slowest processor, if workload is not well-balanced or heterogeneous clusters are used. Asynchronous iteration is a way to mitigate such problems, but most of the works done so far are for shared memory systems. In this paper, we suggest and implement a parallel large sparse linear system solver that improves performance on distributed memory systems like clusters by reducing processor idle times as much as possible by asynchronous iterations.

• The Journal of the Acoustical Society of Korea
• /
• v.21 no.6
• /
• pp.522-530
• /
• 2002

This paper introduces a neural network (NN) -based nonlinear predictor for the LP (Linear Prediction) residual. To evaluate the effectiveness of the NN-based nonlinear predictor for LP-residual, we first compared the average prediction gain of the linear long-term predictor with that of the NN-based nonlinear long-term predictor. Then, the effects on the quantization noise of the nonlinear prediction residuals were investigated for the NN-based nonlinear predictor A new NN predictor takes into consideration not only prediction error but also quantization effects. To increase robustness against the quantization noise of the nonlinear prediction residual, a constrained back propagation learning algorithm, which satisfies a Kuhn-Tucker inequality condition is proposed. Experimental results indicate that the prediction gain of the proposed NN predictor was not seriously decreased even when the constrained optimization algorithm was employed.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.17 no.2
• /
• pp.620-626
• /
• 2016

Estimating the nonlinear seismic response of structure in earthquake engineering is important. Nonlinear static analysis is a typical method, and a variety of methods and techniques for estimating the stiffness of structural system at a certain analysis stage have been introduced and used in numerical structural analysis. On the other hand, such methods have many difficulties in practical usage because they use time-consuming iterative methods or simplified algorithms for calculating the structural stiffness at specific points in the time of nonlinear static analysis. For this reason, this study suggests an accurate and effective method for estimating the stiffness of a structure in nonlinear static analysis. For this goal, existing theories of an incremental step-by-step solution was investigated first. Subsequently, an algorithm available for calculating the precise stiffness of a structural system, each element of which has a bilinear material model, was developed based on the investigated methods. Finally, a computer program, sNs, was developed with the algorithm used.

• Journal of the Korean Society of Hazard Mitigation
• /
• v.1 no.1 s.1
• /
• pp.103-115
• /
• 2001

The current procedure to design hydraulic structures in a small basin area is to estimate the probable rainfall depth using rainfall intensity formula. The estimation of probable rainfall depth has many uncertainties inherent with it. However, it has been inevitable to simplify the nonlinearity if the rainfall in practice. This study attend to address a method which can model the nonlinearity in order to derive better rainfall intensity formula for the estimation of probable rainfall depth. The results show that genetic algorithm is more reliable and accurate than trial-and-error method or nonlinear programming technique(Powell's method) in the derivation of the rainfall intensity formula.

• The Journal of the Acoustical Society of Korea
• /
• v.34 no.3
• /
• pp.177-183
• /
• 2015

This paper proposes a robust nonlinear acoustic echo canceller (NAEC) which is effective for modeling the nonlinearity of a speaker module and the long acoustic echo path within a speech communication environment. The proposed NAEC utilizes a sigmoid pre-processor for modeling the speaker nonlinearity and a partitioned block frequnecy-domain adaptive filter for identifying the acoustic echo path with small delay. Simulation results confirmed that the proposed algorithm achieves excellent performance with much lower computational complexity than the previous NAEC.

• Proceedings of the Korean Society of Soil and Groundwater Environment Conference
• /
• 2000.11a
• /
• pp.64-68
• /
• 2000

장기간 관측된 지하수위 자료를 시계열분석 중의 하나인 전이함수 모형(Transfer Function - Noise model)을 이용하여 분석하였다. 일반적으로 전이함수 모형은 입력 변수와 출력변수와의 관계가 선형적일 때 적용이 가능하며, 자료가 시간에 대해 연속적으로 존재해야 하는 제한이 있다. 강수량과 지하수위의 변동은 비선형적인 관계를 가지고 있어 이러한 전이함수 모형을 직접 적용하는데는 어려움이 있다. 이러한 비선형성의 정도를 감소시키기 위해 물리모형(HYDRUS)을 이용하여 침투량을 계산하고 이를 입력변수로 사용하여 전이함수 모형을 적용하였다. 침투량을 입력변수로 모형을 추정하였을 때, 강수량을 직접 입력자료로 사용했을 경우보다 ME(mean error), RMSE(root-mean-squre error), MAE(mean absolute error)에서 상대적으로 작은 값을 보여주고 있다. TFN 모형의 모수를 추정하기 위해서 Kalman 필터 알고리즘과 최우추정법(Maximum Likelihood Estimation)을 이용하였다. Kalman 필터 알고리즘을 이용하여 불규칙한 관측주기를 갖는 시계열이나 결측값이 있는 시계열에 대해서도 전이함수 모형을 구하였으며, 이를 통해 결측값에 대한 추정이 가능하였다.

• Proceedings of the Korea Institute of Convergence Signal Processing
• /
• 2006.06a
• /
• pp.33-36
• /
• 2006

본 논문은 에지부분에서 뚜렷한 영상의 복원이 가능하도록 하는 칼라 영상의 비선형 보간 기법에 관한 것이다. 일반적으로 칼라영상을 구성하고 있는 성분중 휘도신호(Y)가 에지(edge)성분에 충실한 정보를 갖고 있다는 점에 착안하여 알고리즘 연구를 수행하였다. 일반적인 선형 보간 방법을 사용할 시 영상에서 고주파 대역의 손실을 일으키므로 영상의 화질 열화가 발생한다. 이를 보완하기 위해 본 논문에서는 비선형 보간법인 에지 방향성 보간 방법을 제안하였다 또한 조밀한 에지 영역에서의 에지 방향성 보간의 단점을 극복하기 위해 선형 보간과 에지방향성 보간 기법의 혼합을 통한 화질 열화 제거 기법을 제안한다.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.15 no.1
• /
• pp.87-91
• /
• 2005

The cascade-correlation (CC) learning algorithm of Fahlman and Lebiere is one of the most influential constructive algorithm in a neural network. Cascading the hidden neurons results in a network that can represent very strong nonlinearities. Although this power is in principle useful, it can be a disadvantage if such strong nonlinearity is not required to solve the problem. 3 models are presented and compared empirically. All of them are based on valiants of the cascade architecture and output neurons weights training of the CC algorithm. Empirical results indicate the followings： (1) In the pattern classification, the model that train only new hidden neuron to output layer connection weights shows the best predictive ability; (2) In the function approximation, the model that removed input-output connection and used sigmoid-linear activation function is better predictability than CasCor algorithm.

• Journal of the Korean Institute of Telematics and Electronics T
• /
• v.36T no.2
• /
• pp.21-31
• /
• 1999

This paper presents a new possibility of calculating optimal control for large scale which consist of non-linear dynamic sub-systems using two level hierarchical structures method. And the proposed method is based on the idea of block pulse transformation to simplify the algorithm and its calculation. This algorithm used an expansion around the equilibrium point of the system to fix the second and higher order terms. These terms are compensated for iteratively at the second level by providing a prediction for the states and controls which form of a part of the higher order terms. In this new approach the quadratic penalty terms are not used in the cost function. This allows convergence over a longer time horizon and also provides faster convergence. And the method is applied to the problem of optimization of the synchronous machine. Results show that the new approach is superior to conventional numerical method or other previous algorithm.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.11 no.12
• /
• pp.2382-2390
• /
• 2007

A new polynomial-based predistortion method for linearizing nonlinear power amplifier is proposed. The proposed method finds the predistortion parameter directly without the help of postdistorter whereas most existing polynomial-based predistortion methods calculate the predistortion parameter indirectly from the prostdistorter. First, a new predistortion algorithm is derived based on the assumption that the characteristic of the amplifier is modeled by piecewise linear function. Then it is modified into a proposed method which does not require any assumption or prior knowledge of the amplifier. The proposed method is derived based on the RLS (recursive least squares) algorithm. The proposed technique is simpler to implement than the existing methods and the computer simulation demonstrates that the proposed method is more robust to the initial condition and the saturation region of the amplifier.