• Bulletin of the Society of Naval Architects of Korea
• /
• v.31 no.1
• /
• pp.39-42
• /
• 1994

인공 신경망은 빠른 속도와 안정성 등의 많은 장점을 갖고 있기 때문에 최근 들어서 여러 분야 에서 그 응용이 활발히 연구되고 있다. 인공 신경망의 한 모델인 홉필드 네트워크는 네트워크의 에너지를 최소화시키는 방향으로 네트워크의 상태를 바꾸며, 최소 에너지 상태에서 안정 상태를 유지하는 특징을 갖고 있다. 이러한 흡필드 네트워크의 특징은 흡필드 네트워크를 최적화 문 제에 적용시킬 수 있는 가능성을 제시하고 있다. 기존의 최적화 기법은 기본적으로 국부적인 탐색 기법을 사용하기 때문에, 전역적 최적해를 구하기 위해 초기점을 달리하여 여러번 계산을 수행하여 그 중 가장 좋은 결과를 취하는 방법을 사용하여야 한다. 따라서 이러한 방법은 초 기점의 선택이 결과에 큰 영향을 미치게 되는데, 설계 변수가 많고 제한 조건이 복잡할 경우 초기점 선택에 어려움이 따른다. 본 연구에서는 흡필드 네트워크와 시뮬레이티드 어닐링을 결 합하여 전역적 최적해를 찾는 기법으로서 뉴드-옵티마이저 모델을 제시하고 있다.

• Journal of the Korea Society for Simulation
• /
• v.25 no.1
• /
• pp.1-9
• /
• 2016

This study focuses on the air to air missile control fin planform optimization for the minimizing hinge moment with the considering phenomena of fluid and structure simultaneously. The fluid-structure interaction method is applied for the fluid and structure phenomena simulation of the control fins. A transient-loosely coupled method is used for the fluid-structure interaction simulation because it is suited for using each fluid and structure dedicated simulation software. Searching global optimization point is required many re-calculation therefore in this study, a mathematical model is applied for rapidly calculation. The face centered central composite method is used for generating design points and the 2nd polynomial response surface is sued for generating mathematical model. Global optimization is performed by using the generic algorithm. An objective function is the minimizing travel distance of the center of pressure between Mach 0.7 and 2.0 condition. Finally, the objective function of optimized planform is reduced 7.5% than the baseline planform with satisfying constrained conditions.

• The KIPS Transactions:PartB
• /
• v.14B no.7
• /
• pp.523-530
• /
• 2007

This paper proposes a new stochastic optimization algorithm for hidden Markov models (HMMs) used as a recognizer of automatic lipreading. The proposed method combines a global stochastic optimization method, the simulated annealing technique, and the local optimization method, which produces fast convergence and good solution quality. We mathematically show that the proposed algorithm converges to the global optimum. Experimental results show that training HMMs by the method yields better lipreading performance compared to the conventional training methods based on local optimization.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2006.05a
• /
• pp.305-312
• /
• 2006

K-means clustering algorithm 에서 주로 이루어지는 랜덤 초기화 (random initialization) 방법은 전역 최적화된 해(global minimum)를 찾아내기에 문제점을 지니고 있다. 즉, 여러 횟수의 알고리듬 반복(iteration)을 실행하더라도 전역 최적화된 해를 찾아내기가 매우 힘들며 주어진 자료의 크기(data size)가 큰 경우에 있어서 이는 거의 불가능하다. 본 논문은 이러한 문제점들을 극복하기 위한 방안으로, wavelet을 이용하여 최적의 초기 군집 중심점(initial clustering center)들을 선택하는 방법을 제시한다. 즉, 웨이블릿을 이용한 효과적인 초기화 (initialization)를 통해서 작은 알고리듬 반복 횟수만으로도 전역 최적화에 도달하는 초기화 방법을 기술한다. 이런 초기화 방법이 군집 알고리즘에 사용될 경우, 온라인상에서 실시간 이루어지는 군집 분석에 큰 도움이 된 수 있다.

• Proceedings of the KSME Conference
• /
• 2000.04a
• /
• pp.725-730
• /
• 2000

Genetic algorithms based on the theory of natural selection, have been applied to many different fields, and have proven to be relatively robust means to search for global optimum and handle discontinuous or even discrete data. Genetic algorithms are widely used for unconstrained optimization problems. However, their application to constrained optimization problems remains unsettled. The most prevalent technique for coping with infeasible solutions is to penalize a population member for constraint violation. But, the weighting of a penalty for a particular problem constraint is usually determined in the heuristic way. Therefore this paper proposes, the effective technique for handling constraints, the ranking penalty method and hybrid genetic algorithms. And this paper proposes dynamic mutation tate to maintain the diversity in population. The effectiveness of the proposed algorithm is tested on several test problems and results are discussed.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.8 no.5
• /
• pp.98-106
• /
• 1998

This paper proposes an efficient learning algorithm for improving the training performance of the neural network. The proposed method improves the training performance by applying the backpropagation algorithm of a global optimization method which is a hybrid of a stochastic approximation and a conjugate gradient method. The approximate initial point for f a ~gtl obal optimization is estimated first by applying the stochastic approximation, and then the conjugate gradient method, which is the fast gradient descent method, is applied for a high speed optimization. The proposed method has been applied to the parity checking and the pattern classification, and the simulation results show that the performance of the proposed method is superior to those of the conventional backpropagation and the backpropagation algorithm which is a hyhrid of the stochastic approximation and steepest descent method.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.8 no.4
• /
• pp.27-35
• /
• 1998

Navigation is a method to direct a mobile robot without collision when traversing the environment. This is to reach a destination without getting lost. In this paper, global and local path planning in fixed obstacle and moving obstacle using genetic algorithm are presented. First, mobile robot searches optimal global path using genetic algorithm without falling into local minima. Then if it finds a unknown obstacle, it searches new path without crashing obstacle. Also if there is a moving obstacle, mobile robot searches new optimal path without colliding with the obstacles. Various simulation results show the proposed algorithm can search a shortest path effectively.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.38 no.6
• /
• pp.557-563
• /
• 2010

This paper addresses the issue of adaptive global optimization using Kriging metamodel known as EGO(Efficient Global Optimization). The algorithm adaptively chooses where to generate subsequent samples based on an explicit trade-off between reduction of global uncertainty and exploration of the region of the interest. A strategy that saves the computational cost by using expectations derived from probabilistic nature of approximate model is proposed. At every iteration, a candidate test point that seems to be feasible/inactive or has little possibility to improve for minimum is identified and excluded from updating approximate models. By doing that the computational cost is saved without loss of accuracy.

• Annual Conference on Human and Language Technology
• /
• 2004.10d
• /
• pp.173-180
• /
• 2004

생의학 도메인에서 약어 중의성 해결이란 생의학 문서에 나타난 약어의 원래 형태(long form)를 판별하는 작업이다. 본 논문은 생의학 도메인에서 약어 중의성 해결에 적합한 자질들을 실험적으로 탐색하는데 목적이 있다. 이를 위해서 약어 중의성 해결에 사용할 문맥을 전역 문맥(topical context)과 지역 문맥(local context)으로 구분하고, 각각의 문맥에서 스테밍(stemming), 불용어 제거, 품사 부착 등의 과정을 통해서 다양한 자질들을 고려하도록 한다. 생의학 도메인에서 약어 중의성 해결을 위한 실험 자료의 부족을 해결하기 위해서, 학습 자료와 평가 자료를 자동으로 구축했으며, 평가를 위한 약어로는 기존 연구에서 사용된 두 가지 약어 목록을 사용했다. 또한 단순 베이지언 모델(Naive Bayesian Model)을 이용해서 각 자질들의 유용성을 평가하였다 실험 결과, 전역 문맥이 지역 문맥보다 더 좋은 성능을 보였으며, 전역 문맥에서는 불용어만을 제거한 경우가 각각의 평가 자료에서 94.2%와 96.2%로 가장 좋은 결과를 보였으며, 전역 문맥과 지역 문맥을 함께 사용하는 경우에 각각의 평가 자료에서 1.8%와 0.3%의 성능 향상이 있었다.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.11 no.6
• /
• pp.538-542
• /
• 2001

DNA computing has been applied to the problem of getting an optimal soluting since Adleman's experiment. DNA computing uses strings with various length and four-type bases that makes more useful for finding a global optimal solutions of the complex multi-modal problems This paper presents DNA coding method finding optimal solution of the multi-modal function and compares the efficiency of this method with the genetic algorithms(GA). GA searches efffectively an optimal solution via the artificial evolution of individual group of binary string and DNA coding method uses DNA molecules and four-type bases denoted by the A(Ademine) C(Gytosine);G(Guanine)and T(Thymine). The selection, crossover, mutation operators are applied to both DNA coding algorithm and genetic algorithms and the comparison has been performed. The results show that the DNA based algorithm performs better than GA.