• Journal of the Korean Society of Hazard Mitigation
• /
• v.1 no.3 s.3
• /
• pp.73-82
• /
• 2001

This study examines the applicability of the Nash model and the Diskin model, which are linear and nonlinear runoff models, respectively, by applying optimization techniques to the parameter calibration of the two models. Nonlinear programming which is one of traditional optimization techniques and Genetic Algorithm which has been actively applied recently are used in this study. The Nash and Diskin models which use the calibrated parameter with a flood events are applied to a different flood event in Soyang Dam basin. The results obtained from the parameter calibration show slight discrepancy depending upon the flood events. It has been found in the comparion between the observed hydrograph and the hydrographs obtained from the parameter calibration that the Diskin model can better simulate the observed hydrograph than the Nash model can, especially, for the peak flow. This can be analyzed that the Diskin model which is a nonlinear runoff model is better off in simulating the nonlinear characteristic of the rainfall-runoff process.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.33 no.1C
• /
• pp.12-18
• /
• 2008

This paper This paper proposes a new inverse-quantization(IQ) table interpolation algorithm, specialized Digital Signal Processor(DSP) instructions and hardware architecture for digital audio codecs. Non-linear inverse quantization algorithm is representatively used in both MPEG-1 Layer-3 and MPEG-2/4 Advanced Audio Coding(AAC). The proposed instructions are optimized for the non-linear inverse quantization. The proposed algorithm can minimize operational complexity which reduces total computational load. Performance comparisons show a significant improvement of average error. The proposed instructions and hardware architecture can reduce 20% of the instruction counts and minimize computational loads of IQ algorithms effectively compared with existing IQ table interpolation algorithms. Proposed algorithm can implement commercial DSPs.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2019.11a
• /
• pp.110-113
• /
• 2019

본 논문에서는 임베디드 시스템에서의 양자화 기계학습을 수행할 경우 발생하는 양자화 오차를 효과적으로 보상하기 위한 방법론을 제안한다. 경사 도함수(Gradient)를 사용하는 기계학습이나 비선형 신호처리 알고리즘에서 양자화 오차는 경사 도함수의 조기 소산(Early Vanishing Gradient)을 야기하여 전체적인 알고리즘의 성능 하락을 가져온다. 이를 보상하기 위하여 경사 도함수의 최대 성분에 대하여 직교하는 방향의 보상 탐색 벡터를 유도하여 양자화 오차로 인한 성능 하락을 보상하도록 한다. 또한, 기존의 고정 학습률 대신, 내부 순환(Inner Loop) 없는 비선형 최적화 알고리즘에 기반한 적응형 학습률 결정 알고리즘을 제안한다. 실험결과 제안한 방식의 알고리즘을 비선형 최적화 문제에 적용할 시 양자화 오차로 인한 성능 하락을 최소화시킬 수 있음을 확인하였다.

• The KIPS Transactions:PartB
• /
• v.8B no.2
• /
• pp.130-136
• /
• 2001

본 논문에서는 새로운 학습알고리즘의 비선형 주요성분분석 신경망을 이용한 영상데이터의 효율적인 특징추출에 대하여 제안한다. 제안된 학습알고리즘에서는 최적해로 수렴하는 과정에서 발생할 수도 있는 진동을 억제하여 빠른 속도의 수렴이 가능하도록 하기 위해 모멘트를 이용하였고, 국소최적해를 만났을 때 이를 벗어난 전역최적해로의 수렴을 위한 새로운 연결가중치의 설정을 위하여 동적터널링을 이용함으로써 빠른 수렴속도로 전역최적해에 수렴되도록 학습시킬 수 있다. 제안된 학습알고리즘을 이용한 신경망을 256$\times$256 픽셀의 간암영상과 128$\times$128 픽셀의 얼굴영상을 대상으로 실험한 결과, 기울기하강의 학습알고리즘을 이용한 기존 비선형 주요성분분석 신경망보다 우수한 수렴성능과 특징추출성능이 있음을 확인 할 수 있었다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2008.11a
• /
• pp.27-29
• /
• 2008

시스템 분석에 주로 사용하는 자료 중에는 비선형 자료와 시계열 등이 있다. 이들 자료는 그 함축적인 관계가 매우 복잡하여 전통적인 통계분석 도구로 분석하는데 어려움이 많다. 본 연구에서는 현실 세계에서 다양하게 나타나는 복잡성을 다루기 위하여 하이브리드 진화 신경망 모델링 접근 방법으로 자료를 모형화 하고 이를 통한 학습의 적합도를 살펴본다. 비선형 자료 등을 모형화하기 위한 학습은 역전파 신경망 기법을 이용한다. 학습의 효율을 높이기 의해서 격자감소 학습 알고리즘과 함께 이용하는 유전자 알고리즘은 네트워크 구조를 최적화 시킬 수 있는 초기가중값을 이용한 전역 최소값을 찾는데 이용한다. 학습 결과를 통해 제안된 하이브리드형 접근방법의 학습이 보다 효율적임을 살펴보기 위하여 유전자 알고리즘으로 최적화된 신경망 학습 알고리즘을 비선형 모의자료의 학습에 적용하여 보았다.

• Proceedings of the KIEE Conference
• /
• 2007.07a
• /
• pp.1732-1733
• /
• 2007

본 연구는 지능알고리즘을 이용하여 비선형궤환을 구현하여 비선형시스템을 선형제어이론으로 제어할 수 있는 가능성을 제시한다. 기존의 비선형 궤환 선형화 이론은 비선형계통에 대한 정확한 모델링을 바탕으로 선형화기법을 적용하여 선형제어이론의 적용을 가능케 하는 것이었으나 본연구는 가상의 선형시스템과 SVM을 사용하여 동특성을 알려지지 않은 계통에 대해서도 적용시킬 수 있는 비선형 궤환선형화 기법의 가능성을 제공한다.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.7 no.7
• /
• pp.1462-1470
• /
• 2003

According to the necessity about information security as well as the advance of IT system and the spread of the Internet, a variety of cryptography algorithms are being developed and put to practical use. In addition the technique about cryptography attack also is advanced, and the algorithms which are strong against its attack are being studied. If the linear transformation matrix in the block cipher algorithm such as Substitution Permutation Networks(SPN) produces the Maximum Distance Separable(MDS) code, it has strong characteristics against the differential attack and linear attack. In this paper, we propose a new algorithm which cm estimate that the linear transformation matrix produces the MDS code. The elements of input code of linear transformation matrix over GF$({2_n})$ can be interpreted as variables. One of variables is transformed as an algebraic formula with the other variables, with applying the formula to the matrix the variables are eliminated one by one. If the number of variables is 1 and the all of coefficient of variable is non zero, then the linear transformation matrix produces the MDS code. The proposed algorithm reduces the calculation time greatly by diminishing the number of multiply and reciprocal operation compared with the conventional algorithm which is designed to know whether the every square submatrix is nonsingular.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.15 no.12
• /
• pp.2521-2526
• /
• 2011

This paper proposes a speaker recognition algorithm using a perceptron neural network and LPC (Linear Predictive Coding) cepstrum coefficients. The proposed algorithm first detects the voiced sections at each frame. Then, the LPC cepstrum coefficients which have speaker characteristics are obtained by the linear predictive analysis for the detected voiced sections. To classify the obtained LPC cepstrum coefficients, a neural network is trained using the LPC cepstrum coefficients. In this experiment, the performance of the proposed algorithm was evaluated using the speech recognition rates based on the LPC cepstrum coefficients and the neural network.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.29 no.2B
• /
• pp.163-171
• /
• 2009

Classical linear models have been generally used to analyze and forecast hydrologic time series. However, there is growing evidence of nonlinear structure in natural phenomena and hydrologic time series associated with their patterns and fluctuations. Therefore, the classical linear techniques for time series analysis and forecasting may not be appropriate for nonlinear processes. In recent, the BDS (Brock-Dechert-Scheinkman) statistic instead of conventional techniques has been used for detecting nonlinearity of time series. The BDS statistic was derived from the statistical properties of the correlation integral which is used to analyze chaotic system and has been effectively used for distinguishing nonlinear structure in dynamic system from random structures. DVS (Deterministic Versus Stochastic) algorithm has been used for detecting chaos and stochastic systems and for forecasting of chaotic system. This study showed the DVS algorithm can be also used for detecting nonlinearity of the time series. In this study, the stochastic and hydrologic time series are analyzed to detect their nonlinearity. The linear and nonlinear stochastic time series generated from ARMA and TAR (Threshold Auto Regressive) models, a daily streamflow at St. Johns river near Cocoa, Florida, USA and Great Salt Lake Volume (GSL) data, Utah, USA are analyzed, daily inflow series of Soyang dam and the results are compared. The results showed the BDS statistic is a powerful tool for distinguishing between linearity and nonlinearity of the time series and DVS plot can be also effectively used for distinguishing the nonlinearity of the time series.

• Journal of Korean Society of Transportation
• /
• v.25 no.2 s.95
• /
• pp.157-165
• /
• 2007

In this research, a vertical alignment analysis algorithm was developed. The developed algorithm used acquired data from a vehicle with multiple sensors such as a global positioning system (GPS) an inertial navigation system (INS), and a distance measuring unit (DMI) to collect information about vehicle position and altitude. The vertical alignment analysis algorithm includes the identification of vertical tangent sections, the beginning and ending points of vertical curves, and the calculation of length of vortical curves. Also, the algorithm can help build models for vertical tangent sections and vertical curve sections. In order to verify the algorithm, a field survey was conducted at an actual highway section and the result of the field survey was compared to a highway CAD drawing.