• Journal of the Korean Institute of Intelligent Systems
• /
• v.5 no.3
• /
• pp.83-91
• /
• 1995

A neural network with one or more layers of hidden units can be trained using the well-known error back propagation algorithm. According to this algorithm, the synaptic weights of the network are updated during the training by propagating back the error between the expected output and the output provided by the network. However, the error back propagation algorithm is characterized by slow convergence and the time required for training and, in some situation, can be trapped in local minima. A theoretical formulation of a new fast learning method based on tabu search method is presented in this paper. In contrast to the conventional back propagation algorithm which is based solely on the modification of connecting weights of the network by trial and error, the present method involves the calculation of the optimum weights of neural network. The effectiveness and versatility of the present method are verified by the XOR problem. The present method excels in accuracy compared to that of the conventional method of fixed values.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.27 no.6A
• /
• pp.517-523
• /
• 2002

The closed from of sums and weighted sums of the score functions of the locally optimum rank detectors are obtained in this paper. When we consider the asymptotic performance characteristics of a detector based on rank and sign statistics, the sums and weighted sums of the score functions have to be prepared. The efficacy of a detector can be obtained from the sums and weighted sums of the score functions. Score functions based on rank statistics, as well as those based on magnitude rank and sign statistics, have also been considered, which includes most score functions presented in the literature.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.31 no.10C
• /
• pp.973-980
• /
• 2006

In most of the studies on locally optimum detection assumes independent observations. The use of an independent observation model may cause a considerable performance degradation in detection applications of modern high data rate communication systems exhibiting dependence among interference components. In this paper, we address the detection of weak known signals in multiplicative and first order Markov additive noises. In Part 1, the test statistics of the locally optimum detectors are investigated in detail. In Part 2, the asymptotic and finite sample-size performance of several detectors are obtained and compared, confirming that the dependence among interference components need to be taken into account to maintain performance appropriately.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.30 no.6C
• /
• pp.532-540
• /
• 2005

In this paper, a sequential detection scheme is proposed as a combination of a novel weak-signal and a locally optimum(LO) detection schemes. In Part 1, we propose a novel sequential detection scheme for weak signals and show some interesting threshold properties and examples. In Part 2, the performance of the proposed sequential detection scheme is compared with that of the fixed sample size(FSS) test, sequential probability ratio test (SPRT), and truncated sequential probability ratio test(TSPRT).

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.36S no.12
• /
• pp.51-57
• /
• 1999

In this paper, an adaptive thresholding of the nonuniformly contrasted images is proposed through using the contrast pre-enhancement of the local regions and the bilinear interpolation between the local threshold values. The nonuniformly contrasted image is decomposed into 9${\times}$9 sized local regions, and the contrast is enhanced by intensifying the gray level difference of each low contrasted or blurred region. Optimal threshold values are obtained by iterative method from the gray level distribution of each contrast-enhanced local region. Discontinuities are reduced at the region of interest or at the characters by using bilinear interpolation between the neighboring threshold surfaces. Character recognition experiments are conducted using backpropagation neural network on the characters extracted from the nonuniformly contrasted document, PCB, and wafer images binarized through using the proposed thresholding and the conventional thresholding methods, and the results prove the relative effectiveness of the proposed scheme.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2002.12a
• /
• pp.263-266
• /
• 2002

데이터의 군집화를 수행할 때 최적 군집수 결정은 군집 결과의 성능에 많은 영향을 미친다. 특히 K-means 방법에서는 초기 군집수 K에 따라 군집결과의 성능 차이가 많이 나타난다. 하지만 대다수의 군집분석에서 초기 군집수의 결정은 경험을 바탕으로 하여 주관적으로 결정된다. 이때 개체수와 속성수가 증가하면 이러한 결정은 더욱 어려워지며 이때 결정된 군집수가 최적이 된다는 보장도 없다. 본 논문에서는 군집의 수를 자동으로 결정하고 그 결과의 유효성을 보장하기 위해 유전자 알고리즘에 기반한 최적 군집수 결정 방안을 제안한다. 데이터의 속성에 근거한 초기 해 집단이 생성되고, 해 집단 내에서 최적화된 군집수를 찾기 위해 교차 연산이 이루어진다. 적합도 값은 전체 군집화의 비 유사성의 합의 역으로 결정되어 전체적인 군집화 성능이 향상되는 방향으로 수렴된다. 또한 지역 국소값을 해결하기 위해 돌연변이 연산이 사용된다. 그리고 유전자 알고리즘의 학습 시간의 비용을 줄이기 위해 붓스트랩 기법이 적용된다.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.44 no.7
• /
• pp.620-628
• /
• 2016

The homotopy algorithm provides a robust method for determining optimal control, in some cases the global minimum solution, as a continuation parameter is varied gradually to regulate the contributions of the nonlinear terms. In this paper, the Successive Backward Sweep (SBS) method, which is insensitive to initial guess, augmented with a homotopy algorithm is suggested. This approach is effective for highly nonlinear problems such as low-thrust trajectory optimization. Often, these highly nonlinear problems have multiple local minima. In this case, the SBS-homotopy method enables one to steadily seek a global minimum.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.16 no.11
• /
• pp.1125-1131
• /
• 1991

In a generalized model with which we can represent multiplicative noise as well as purely additive noise, we consider detection of composite signals which contain both deterministic and stochastic signal components. To illustrate the performance of the locally optimum detectors, finite sample-size performance characteristics are obtained and compared with those of other detectors.

• Transactions of the Korean Society of Mechanical Engineers
• /
• v.15 no.5
• /
• pp.1706-1716
• /
• 1991

본 연구에서는 혼합대류하는 연직원통휜 열전달에 미치는 복사효과에 Rossel- and 근사해법을 도입한 층류 경계층방정식과 휜 전도방정식을 동시에 해석하여 복사 -전도매개변수(radiation-conduction parameter) M, 대류-전도 매개변수(convectionc- onduction parameter) N$_{c}$, 부력매개변수(buoyancy force parameter) R$_{i}$ 그 리고 횡곡률 매개변수(transverse curvature parameter) .lambda.를 계산 파라미터로 하여 휜 최적설계의 기본이 되는 총열전단률, 무차원 국소등가열전달계수, 무차원 국소열플 럭스 및 휜온도분포를 계산할 수 있는 일반성있는 algorithm을 개발하고 휜 성능에 관 한 결과들을 제공함으로써, 좀 더 완전한 복사 열플럭스를 도입한, 다음 연구를 위한 범용의 기초자료를 얻는데 본 연구의 목적이 있다.

• The KIPS Transactions:PartB
• /
• v.8B no.4
• /
• pp.327-334
• /
• 2001

본 논문에서는 기울기하강과 동적터널링이 조합된 학습알고리즘의 다층신경망을 이용한 고신회성의 회귀분석 모델을 제안하였다. 기울기하강은 빠른 수렴속도의 최적화가 가능하도록 하기 위함이고, 동적터널링은 국소최적해를 만났을 때 이를 벗어난 새로운 연결가중치를 설정하여 전역최적해로 수렴되도록 하기 위함이다. 또한 대용량의 입력 데이터를 통계적으로 독립인 특징들의 집합으로 변환시키는 주요성분분석 기법의 속성을 살려 학습데이터의 차원을 감소시킴으로서 고차원의 학습데이터에 따른 회귀분석 모델의 제약도 동시에 해결하였다. 제안된 기법의 신경망을 3개의 독립변수 패턴을 가진 암모니아 제조공정문제와 10개의 독립변수 패턴을 가진 자동차 연비문제에 각각 적용하여 시뮬레이션한 결과, 기존의 역전과 알고리즘의 신경망이나 주요성분분석에 의한 차원을 감소시키지 않은 학습패턴을 이용한 신경망보다 각각 더욱 우수한 학습성능과 회귀성능이 있음을 확인할 수 있었다. 또한 학습패턴의 영평균 정규화로 회귀용 신경망의 성능을 더욱 더 개선하였다.